@0xlelouch_ Avatar @0xlelouch_ Abhishek Singh

Abhishek Singh posts on X about redis, ai, if you, systems the most. They currently have [------] followers and [---] posts still getting attention that total [-------] engagements in the last [--] hours.

Engagements: [-------] #

Engagements Line Chart

Mentions: [--] #

Mentions Line Chart

Followers: [------] #

Followers Line Chart

CreatorRank: [------] #

CreatorRank Line Chart

Social Influence

Social category influence technology brands 15.2% stocks 11.27% finance 8.33% countries 4.41% social networks 4.41% cryptocurrencies 1.47% travel destinations 0.98% gaming 0.49% luxury brands 0.49% financial services 0.49%

Social topic influence redis #1, ai 7.84%, if you 7.35%, systems #1382, clickhouse 5.39%, staff #552, data 3.43%, events #1171, in the 3.43%, fine 3.43%

Top accounts mentioned or mentioned by @omranchi00 @devxritesh @consciousride @grok @akashbitm787 @umesh__digital @patilvishi @rahulbagal__ @treeapostle @codewiddan @pragyan2214 @_adithya_n_g @rahulbagal @iamgoroot @navneet_rabdiya @ashutosh_1205 @utkarshmarathe @egorseno7 @lalitkhu @umeshdigital

Top assets mentioned Alphabet Inc Class A (GOOGL) Uber Technologies, Inc. (UBER) Microsoft Corp. (MSFT) Cloudflare, Inc. (NET)

Top Social Posts

Top posts by engagements in the last [--] hours

"As a Backend Engineer in [----] aiming for Staff please learn: [--]. One language deeply (Go/Rust/Java) Not I can write APIs but runtime model memory concurrency profiling GC behavior (if any) and how to read stack traces like a native. [--]. Data modeling and storage fundamentals Relational modeling constraints isolation levels indexes query plans locks deadlocks migrations backup/restore partitioning. Most scaling problems are schema + query shape problems. [--]. Distributed systems basics that actually show up in prod Consistency vs availability timeouts retries idempotency backpressure message"
X Link 2026-01-21T09:37Z 17.3K followers, 553.2K engagements

"Your load balancer uses round-robin to distribute traffic across [--] servers. One server has 2x CPU capacity but it gets the same traffic as weaker servers and is underutilized. How will you optimize load distribution"
X Link 2026-02-14T09:45Z 17.3K followers, 16.6K engagements

"As a backend guy who has worked with payments before I can take this up. Amazon Prime Day sale is crazy scale. We need to handle quite a few things to properly design this system. The client (FE or Amazon Android app) sends an idempotency key with each checkout attempt. The server stores this key in a fast DB like Redis with a TTL of 24h+. It's better to attach states to payments: CREATED AUTHORIZED CAPTURED SETTLED (and FAILED/CANCELLED). The key idea is that state only moves forward so we don't need to worry about duplicates - thanks to the client-sent idempotency key. We pass the same"
X Link 2026-02-10T17:17Z 17.3K followers, 48.9K engagements

"Users' shopping carts keep getting lost because each request goes to a different server and cart data isn't shared. How will you maintain session state across servers"
X Link 2026-02-14T14:04Z 17.3K followers, 12.3K engagements

"Use local rate limiting with eventual consistency. Each gateway instance tracks limits locally in memory syncs periodically to Redis. Implement sliding window in Redis (sorted sets). Or use token bucket algorithm per gateway instance. Accept slight over-limit during bursts (96-104 instead of exact 100). Use Redis Cluster to shard rate limit counters. Alternative: distributed rate limiting with Envoy/Istio. For extreme scale: approximate counting (HyperLogLog). Add rate limit headers in response. Pre-allocate quota to each gateway instance. Your API Gateway becomes a bottleneck at 100K"
X Link 2026-02-15T09:15Z 17.3K followers, [----] engagements

"We hired a senior who had Google + Amazon on his resume. His employment pattern was weird: [--]. Big tech. [--]. [--] months. [--]. Quit. [--]. Repeat. In interviews he crushed DSA. Clean solutions in 15-20 mins. In real work: [--]. Couldnt collaborate. Every PR turned into a fight. [--]. No ownership. Prod issue at [--] AM meant not my service. [--]. Didnt debug. Asked for logs didnt read them. [--]. Never did the boring stuff: runbooks alerts oncall fixes. He got fired (and luckily found another senior role at another startup). So DSA might get you in (although this is changing a lot nowadays). But teamwork + ownership"
X Link 2026-02-10T16:30Z 17.3K followers, 623.5K engagements

"I've experienced this many times in my life especially in the context of both my career and learning new skills. If you are just doing something consistently. You win. Why Because just the sheer number of consistent individuals are very very few. So it's pretty crazy. Do it scared. Do it imperfect. Do it alone. Just do it. Do it scared. Do it tired. Do it doubting. Do it imperfect. Do it alone. Do it anyway. Do it scared. Do it tired. Do it doubting. Do it imperfect. Do it alone. Do it anyway"
X Link 2026-02-15T20:43Z 17.3K followers, [----] engagements

"Your API Gateway becomes a bottleneck at 100K requests/sec. Adding more gateway instances doesn't help because they all hit the same rate-limiting database. How will you scale rate limiting"
X Link 2026-02-15T20:48Z 17.3K followers, 10.2K engagements

"As a senior backend engineer please learn: [--]. SOLID design principles [--]. Multithreading and concurrency models [--]. Immutability and safe state management [--]. Streaming and messaging systems (Kafka queues pub/sub) [--]. Caching strategies (in-memory distributed eviction policies) [--]. Security fundamentals (TLS/SSL JWT OAuth2 authentication vs authorization) [--]. Core design patterns (factory decorator singleton observer) [--]. Test-Driven Development (TDD) and testability [--]. API design and versioning (REST gRPC backward compatibility) [--]. Database fundamentals (schema design indexing transactions) 11."
X Link 2026-01-04T12:02Z 17.3K followers, 361.1K engagements

"As a Backend Engineer in [----] please learn: One backend language deeply (Go/Java/Rust/Python) not just syntax but profiling concurrency memory and stdlib patterns. HTTP fundamentals methods headers caching content negotiation ETags idempotency and real-world pagination. API design REST done right versioning backward compatibility error models and sane request/response contracts. Databases Postgres/MySQL internals basics indexes query plans transactions isolation levels locks and connection pooling. Caching Redis patterns TTLs invalidation strategy stampede protection hot keys and when not to"
X Link 2026-01-21T18:07Z 17.3K followers, 344.7K engagements

"20 Most asked Golang LLD Interview problems [--]. Design an In-Memory Cache (TTL eviction thread-safe) [--]. Design a Rate Limiter (token bucket / leaky bucket per-key limits) [--]. Design a Circuit Breaker (closed/open/half-open rolling window stats) [--]. Design a Worker Pool (bounded queue graceful shutdown) [--]. Design a Pub-Sub (topics subscribers backpressure fan-out) [--]. Design a Job Queue (ack retry dead letter queue idempotency) [--]. Design a Pipeline (stages bounded buffers context propagation) [--]. Design a Connection Pool (max conns idle timeout health checks) [--]. Design a Config system (env + file +"
X Link 2026-02-06T09:50Z 17.3K followers, 27.8K engagements

"Uber DSA prep cheatsheet if you only have time for [--] topics: [--]. Two Pointers: [--] (3Sum) [--] (Container With Most Water) [--]. Sliding Window: [--] (Longest Substring Without Repeating) [---] (Longest Repeating Character Replacement) [--]. BFS/DFS on Graphs: [---] (Number of Islands) [---] (Rotting Oranges) [--]. Heaps (Top K): [---] (Top K Frequent Elements) [---] (Kth Largest Element) [--]. Binary Search: [--] (Search in Rotated Sorted Array) [---] (Find Minimum in Rotated Sorted Array) Do these [--] problems timed. If you can solve each in [--] to [--] mins with clean code youre in decent shape. I use these as my "initial"
X Link 2026-02-13T07:10Z 17.3K followers, 44.7K engagements

"twitter bait tiers (for software devs) - pretending to be a 10x dev (but cant merge without breaking prod) - pretending to be a staff engineer (job title: thread writer) - pretending to be ex-FAANG (worked there as a visitor) - pretending to be first world (timezone cosplay + $ signs everywhere) - pretending to be top rank in Codeforces / LeetCode (profile link mysteriously missing) - pretending to be a girl (engagement speedrun) - pretending to be a VC-backed founder (company: stealth for [--] years) - pretending to be an AI wizard (just wrapper APIs + vibes) - pretending to be a designer"
X Link 2026-02-15T11:12Z 17.3K followers, [----] engagements

"You serve a video that goes viral. [--] million users try to download it simultaneously overwhelming your origin server. How will you handle this traffic spike"
X Link 2026-02-16T07:09Z 17.3K followers, [----] engagements

"Which is your favorite or most used design pattern [--]. Event-Driven Architecture (pub/sub) [--]. Saga (distributed transactions) [--]. CQRS (separate read/write models) [--]. Event Sourcing (state from events) [--]. Circuit Breaker (stop cascading failures) [--]. Bulkhead (isolation boundaries) [--]. Outbox + CDC (reliable event publish) [--]. Idempotency (safe retries) [--]. Rate Limiting (protect dependencies) [--]. API Gateway (edge auth routing throttling) https://twitter.com/i/web/status/2023404216392585344 https://twitter.com/i/web/status/2023404216392585344"
X Link 2026-02-16T14:28Z 17.3K followers, [----] engagements

"Age of Databases (with categories) (R = Relational KV = KeyValue DOC = Document COL = Columnar GRAPH = Graph OLAP = Analytics DIST = Distributed SQL EMB = Embedded TS = Time-series) --- 🟥 Oracle DB (R) [--] years (1979) 🟥 Teradata (OLAP) [--] years (1979) 🟥 IBM Db2 (R) [--] years (1983) 🟥 Sybase ASE (R) [--] years (1988) 🟥 Microsoft SQL Server (R) [--] years (1989) 🟥 PostgreSQL (R) [--] years (1996) (roots 86) 🟥 MySQL (R) [--] years (1995) 🟥 MariaDB (R) [--] years (2009) 🟦 CockroachDB (DIST) [--] years (2014) 🟦 TiDB (DIST) [--] years (2016) 🟩 Berkeley DB (KV/EMB) [--] years (1991) 🟩 Redis (KV) [--] years"
X Link 2025-12-08T08:04Z 16.8K followers, 26.9K engagements

"1. Linux kernel is C [--]. Windows kernel core is C [--]. macOS kernel (XNU) is mostly C [--]. UNIX is C [--]. Embedded systems firmware is C [--]. IoT devices run C [--]. Routers and network switches run C [--]. TCP/IP stack is written in C [--]. HTTP servers like Nginx are C [--]. Databases like PostgreSQL are C [--]. SQLite is C [--]. Redis is C [--]. Git is C [--]. Docker core components are C [--]. Kubernetes depends on C libraries at the OS layer [--]. Compilers like GCC are C [--]. LLVM runtime is C/C++ [--]. Operating system bootloaders are C [--]. Aerospace flight software is C [--]. Automotive ECUs run C [--]. Medical devices run"
X Link 2025-12-28T10:32Z 17.2K followers, 632.6K engagements

"1. Kubernetes is Go [--]. Docker is Go [--]. Terraform is Go [--]. Prometheus is Go [--]. Grafana backend is Go [--]. etcd is Go [--]. Consul is Go [--]. Nomad is Go [--]. Istio control plane is Go [--]. Helm is Go [--]. ArgoCD is Go [--]. Flux is Go [--]. InfluxDB is Go [--]. CockroachDB core is Go [--]. Temporal is Go [--]. OpenTelemetry collectors are Go [--]. Cloudflare control systems are Go [--]. Dropbox backend services are Go [--]. Uber backend infrastructure is Go [--]. Twitch backend services are Go [--]. PayPal services are Go [--]. Stripe infrastructure tools are Go [--]. Google cloud networking tools are Go [--]. HashiCorp tooling"
X Link 2025-12-28T10:36Z 16.8K followers, 114.7K engagements

"20MB JSON is not a normal request body consider the design around streaming backpressure and moving work off the hot path. Parse incrementally instead of json. Unmarshal into giant structs: use a streaming decoder (json.Decoder in Go) decode token-by-token / item-by-item and validate as you go so you can fail fast without allocating the whole tree. Add compression (gzip/zstd) to cut bandwidth. For truly heavy ingestion it gets more interesting switch to an async ingest pattern: client uploads to object storage (presigned URL) you accept a small metadata request then a worker parses in the"
X Link 2026-01-10T13:45Z 17K followers, 18.1K engagements

"If you: [--]. Do 2-3 hours of deep work with notifications off and one clear outcome (ship a feature fix a perf issue close a bug cluster) [--]. Write code daily but also read code daily (your codebase open source PRs) and learn how good engineers structure systems [--]. Get strong at debugging: logs metrics traces reproducing locally bisecting writing the smallest failing test then fixing root cause [--]. Learn one core backend topic every month and apply it at work (SQL indexes query plans caching queues concurrency rate limits idempotency timeouts) [--]. Write design notes before big changes (tradeoffs"
X Link 2026-01-22T09:37Z 16.4K followers, 108K engagements

"This was asked in a Google interview recently. You deployed a new feature to production but 30% of users are seeing stale data while 70% see fresh data. How will you debug and fix this issue Very common interview question asked these days: How do you handle token expiry without asking the user to log in again Very common interview question asked these days: How do you handle token expiry without asking the user to log in again"
X Link 2026-01-26T12:22Z 15.9K followers, 556.9K engagements

"Deliveroo-style system design interviews often look simple on paper but theyre really testing ops instincts. A good question to practice: Design order tracking for 50k concurrent orders updates every 12s show ETA on user app + rider app and survive region blips. A solid solution shape: [--]. Write path: rider app sends GPS updates API Gateway Kafka (key = orderId) [--]. Stream processor: enrich (map matching speed traffic) + compute ETA [--]. Read path: materialize to Redis (orderId latest state) + Postgres for history/audit [--]. Fanout: WebSockets/SSE to users + riders with rate limits (ex: 1"
X Link 2026-01-27T23:11Z 15.9K followers, [----] engagements

"A junior got placed at Amazon Hyderabad [--] months back. CTC: [--] LPA (24 base + [--] joining bonus + [--] RSUs). Role: SDE-2. This week he got laid off. Severance is surprisingly decent: [--] weeks base pay encashed leaves full notice payout plus RSUs handled per policy. Roughly [--] to [--] lakh cash runway depending on tax and leaves. He is not even panicking. He said: I basically got paid to job hunt for [--] months. Updating resume calling referrals targeting [--] applications a week and focusing on [--] stacks only. layoffs hurt but he is already looking again. My friend moved to Seattle for Amazon. $170K total"
X Link 2026-01-28T22:15Z 16.5K followers, 896.9K engagements

"If its OLTP Postgres If its simple + embedded SQLite If its massive OLTP scale MySQL If its time-series metrics Prometheus / Timescale If its logs & search OpenSearch / Elasticsearch If its analytics (OLAP) ClickHouse If its big warehouse Snowflake / BigQuery If its real-time analytics pipelines Druid If its caching-as-a-db Redis If its key-value at insane throughput DynamoDB If its wide-column workloads Cassandra If its graph relationships Neo4j If its vectors / RAG pgvector / Milvus / Pinecone If its distributed SQL correctness CockroachDB If its one machine super fast analytics DuckDB"
X Link 2026-01-29T09:51Z 17K followers, [----] engagements

"Layoffs this year (2025 highlights): Intel: 21000+ (20% of workforce) Microsoft: 6500+ (3% of global workforce) Wicresoft (Microsoft JV): [----] impacted (China operations shutting down) CrowdStrike: [---] (5%) Google: hundreds (Platforms & Devices / Google TV cuts also reported) Scale AI: [---] (+ [---] contractors cut) Meta: 100+ (Reality Labs) Amazon: [---] (Devices & Services) Same reminder new year: companies will cut headcount and call it efficiency while the business stays fine. Your move as an engineer: Build the emergency fund. Keep interview-ready. Keep your skills compounding. Dont confuse a"
X Link 2026-01-29T09:58Z 16.9K followers, 44.7K engagements

"Here are the system design questions that keep showing up (and what the interviewer is really probing): [--]. Design a URL Shortener (but at Amazon scale) Theyll push: hot keys read-heavy traffic cache strategy storage growth TTL abuse prevention. [--]. Design a Rate Limiter for an API Gateway Theyll push: token bucket vs leaky bucket per-user vs per-IP distributed counters Redis vs in-memory consistency under bursts. [--]. Design a Notification System (email/SMS/push) Theyll push: retries DLQ idempotency keys delivery guarantees provider failures fan-out at scale. [--]. Design a File Upload / S3-like"
X Link 2026-01-29T10:06Z 16K followers, 47.3K engagements

"You are building an image gallery API where users have uploaded 100K images totaling 50GB. How will you return the gallery to the client efficiently"
X Link 2026-01-29T17:53Z 16.8K followers, 14.6K engagements

"You're inserting [--] billion rows per day into ClickHouse for analytics but queries are timing out after you reach [---] billion total rows. How will you optimize query performance Production issue at Cloudflare https://twitter.com/i/web/status/2017219626820911209 https://twitter.com/i/web/status/2017219626820911209"
X Link 2026-01-30T12:53Z 17K followers, 32.6K engagements

"You are building an analytics dashboard API that needs to return time-series data but the table has 50M data points. How will you return the data to the client efficiently"
X Link 2026-01-30T17:54Z 16.5K followers, [----] engagements

"Golang libraries I refuse to build backend systems without in 2026: 🔶 chi (or gin) = HTTP routing that stays out of your way 🔶 zap (or slog) = structured logging that wont slow prod 🔶 validator/v10 = request validation that isnt if err = nil spaghetti 🔶 sqlc = type-safe SQL without ORM magic 🔶 pgx = Postgres driver that actually performs 🔶 redis/go-redis = caching + rate limits + locks 🔶 grpc-go = service-to-service calls done right 🔶 protobuf = contracts that dont break silently 🔶 go-playground/middleware patterns = timeouts recover request IDs 🔶 prometheus/client_golang = metrics"
X Link 2026-01-30T18:28Z 16.5K followers, 52.4K engagements

"POV: Your AI agent has a better social life than you do. 🦞 Welcome to Molt Book - the front page of the Agent Internet. Its where your Moltbot goes to doomscroll share skills and talk about you behind your back (kidding. mostly). Human status: Read-only. Bot status: Uncaged. https://twitter.com/i/web/status/2017362010581963087 https://twitter.com/i/web/status/2017362010581963087"
X Link 2026-01-30T22:19Z 17.2K followers, [----] engagements

"Do not use UUID as your primary key by default use BIGINT identity for OLTP (smaller indexes faster joins) and add UUID only when you need cross-system uniqueness. Do not use OFFSET pagination on big tables use keyset pagination (WHERE id last_id ORDER BY id LIMIT n) with a stable index. Do not use SELECT * in hot paths use narrow selects (only columns you need) + cover with the right index. Do not use COUNT(*) on a giant table for UX counters use approx counts / cached counters / rollups depending on accuracy needs. Do not use JSONB for everything use proper relational columns for fields you"
X Link 2026-01-31T00:43Z 17K followers, 188.2K engagements

"You're using Redis as a message queue and messages are being processed twice by different workers. How will you ensure exactly-once processing Prod issue we faced in the last startup https://twitter.com/i/web/status/2017563376323547197 https://twitter.com/i/web/status/2017563376323547197"
X Link 2026-01-31T11:39Z 16.4K followers, 79.6K engagements

"Your ClickHouse backup takes [--] hours for a 50TB database and you need faster recovery time. How will you optimize backup/restore strategy Production challenge at Cloudflare https://twitter.com/i/web/status/2017626767100678329 https://twitter.com/i/web/status/2017626767100678329"
X Link 2026-01-31T15:51Z 17K followers, 10.3K engagements

"Two engineers. Similar skills. One gets hired fast one keeps getting ignored. And the difference is not talent. It is trust. The guy who gets hired is not always the best coder in the room. He is the one who makes the hiring manager feel safe. Safe that if he joins tickets will not get stuck. On-call will not be a nightmare. Knowledge will not disappear. New joiners will ramp without begging for help. The system will not collapse when he takes leave. I have literally seen people get hired because they came with a small working system mindset: clear docs clean PRs good commit messages"
X Link 2026-01-31T17:48Z 17.2K followers, [----] engagements

"UUIDv4 vs UUIDv7 vs ULID in practice: [--]. UUIDv4: pure random. Great uniqueness. Terrible for DB indexes. Inserts scatter B-tree bloats cache misses. p99 gets weird at scale. [--]. UUIDv7: time-ordered + random. Keeps globally unique but inserts are mostly sequential. Works nice with Postgres/MySQL primary keys. [--]. ULID: also time-ordered human-friendly sortable as string. Good choice if you need lexicographic order in logs or URLs. Watch clock skew if you generate on many nodes. In Production: default to UUIDv7 for primary keys. Use ULID when you want sortable strings. Avoid v4 for hot indexed"
X Link 2026-01-31T18:22Z 16.4K followers, 100.1K engagements

"A service runs [--] instances behind a load balancer but during deploy 10% of requests fail with [---] errors for [--] minutes. How will you achieve zero-downtime deployments"
X Link 2026-01-31T21:13Z 16.5K followers, 27.1K engagements

"You are implementing a file download API but some files are 5GB in size. How will you return the file to the client efficiently"
X Link 2026-02-01T05:54Z 16.5K followers, 105.2K engagements

"You're streaming real-time events to ClickHouse using Kafka but data appears in queries 5-10 minutes late. How will you reduce this latency Asked at Uber interview https://twitter.com/i/web/status/2017902645177123319 https://twitter.com/i/web/status/2017902645177123319"
X Link 2026-02-01T10:07Z 17K followers, 44.9K engagements

"ClickHouse just raised $400M. Valued at $6B+. 3000+ paying customers. 250% ARR growth. And most backend engineers have never touched it (at least I didn't before last year). This is the database quietly powering the AI era. Here are two success stories that I came across. --- Langfuse is the most popular open-source LLM observability platform. 20K+ GitHub stars. 26M+ SDK installs per month. They were running PostgreSQL as their backend. It worked fine until it didn't. As AI companies started deploying LLMs in production Langfuse's ingestion volumes exploded. Billions of rows. Complex queries"
X Link 2026-02-01T13:14Z 17.2K followers, 27.9K engagements

"Two friends same goal: buy a [---] Cr home in Pune. Friend A works in the USA. $160K salary but after taxes + rent + car + insurance he saves $2.5K/month. He also wires money to India and loses a bit on FX + transfers. Down payment 30L takes [--] years if nothing goes wrong. Friend B works in India. [--] LPA saves 1.2L/month because parents support no $3K rent no surprise medical bills cheaper life. Down payment 30L takes [--] years. Same house in India but the math is different: in the USA you earn more but burn more. In India you earn less but if your expenses stay low you can build the down payment"
X Link 2026-02-01T15:28Z 16.8K followers, 33.4K engagements

"@0xAmbesh Crazy One thing I really like about ClickHouse is that you can tailor it exactly to your needs by simply tuning a few knobs"
X Link 2026-02-01T17:32Z 16.4K followers, [---] engagements

"You have an e-commerce API that needs to return search results but a query like "laptop" matches 500K products. How will you return search results to the client efficiently"
X Link 2026-02-02T05:54Z 16.5K followers, 18.6K engagements

"I am on call this week (Feb 2Feb 9). I will share the issues we faced this week in detail. Issue 2: We had our ClickHouse cluster throwing memory exceeded because our dashboards were doing the classic mistake: SELECT * . LIMIT [--] inside a subquery just to "load everything" and then pick the max from it. Under load it was brutal: 80GB+ memory 25s+ compute queries timing out and dashboards showing [---] internal error for no reason. Initially we just increased the memory for this cluster and some of the queries worked some of the time for some of the clients but it was still not fixed. Switched"
X Link 2026-02-02T15:05Z 16.4K followers, [----] engagements

"You are implementing a social media feed API but each user has 100K+ posts in their timeline. How will you return the feed to the client efficiently"
X Link 2026-02-02T17:55Z 16.9K followers, [----] engagements

"As software engineers we frequently encounter design decisions. Let's break this down into a step-by-step process. [--]. I would pick Amazon SQS + AWS Lambda for this requirement set. [--]. The volume is not Kafka volume: [-----] events/hour is [--] events/sec each 2KB. This is tiny. We do not need Amazon MSK complexity just to move [--] msgs/sec. [--]. The killer constraint is ordering for 30% of event types. We can solve this cleanly by splitting lanes: - Standard SQS queue for the 70% unordered events (cheap + high throughput) - SQS FIFO queue for the 30% ordered events (use MessageGroupId per entity"
X Link 2026-02-02T20:29Z 16.8K followers, [----] engagements

"Your service logs 1TB per day and you can't search logs older than [--] hours because the system is too slow. How will you make logs searchable and cost-effective"
X Link 2026-02-03T17:30Z 17K followers, 84K engagements

"@patilvishi Yes. Or one can simply use datadog 😅"
X Link 2026-02-03T17:52Z 16.9K followers, [----] engagements

"US has engineers making $100K$300K. I landed a remote $120K role early in my career. Dollars hitting an Indian bank account feels unreal for a while. I did it for [--] years: [--]. Saved most of it (kept lifestyle same). [--]. Cleared debt. [--]. Built a 12-month runway in index funds + cash. Then I left for a smaller Indian company with lower pay. People thought I just downgraded. But the fact is I just bought time. Less pressure more impact great career growth and trajectory. So yeah sometimes you just need to be smart enough for 1yr or [--]. When you interview developers in the US earning $100K$300K you"
X Link 2026-02-03T18:35Z 17K followers, 519.7K engagements

"As a Golang backend dev you are expected to set up Alerts + Monitoring too. Not just writing handlers and gRPC methods. In Go interviews (esp backend + platform teams) a very common question is: How do you expose Go service metrics and integrate them with Prometheus + Grafana for monitoring and alerting As a Java backend dev you are expected to configure Alerts & Monitoring Tools like Splunk Dynatrace Grafana ELK. So in springboot interviews it is commonly asked- How can you integrate Spring Boot Actuator with external monitoring and alerting systems such as Grafana As a Java backend dev you"
X Link 2026-02-04T05:27Z 16.9K followers, 26.8K engagements

"My solution: Reduce log verbosity (remove debug logs in prod). Implement structured logging (JSON) for better filtering. Use log sampling (log 1% of success 100% of errors). Send logs to centralized system (ELK Loki Datadog). Create indexes on timestamp and severity. Archive old logs to cold storage (S3) after [--] days. Use log levels appropriately. Add retention policies (30-90 days). Pre-filter before shipping (only ERROR/WARN). Your service logs 1TB per day and you can't search logs older than [--] hours because the system is too slow. How will you make logs searchable and cost-effective Your"
X Link 2026-02-04T08:27Z 17.3K followers, 12.5K engagements

"Kafka vs RabbitMQ vs NATS (in Go) is mostly a decision of what you want to guarantee and what youre willing to build. [--]. Ordering Kafka: ordered per partition. Your key decides the specific partition. RabbitMQ: ordering is best-effort and breaks with requeue/parallel consumers. NATS: core has no persistence; JetStream gives ordering per subject/stream but you still need to design for retries. [--]. Delivery guarantees All three can do at-least-once in practice. Exactly-once is a workflow not a checkbox. In Go: ack only after side effects are committed. [--]. Deduplication None magically saves you."
X Link 2026-02-04T13:20Z 17K followers, 18.7K engagements

"You're using Kafka to stream order events but consumers are falling behind by [--] hours during peak traffic. How will you catch up Production issue at LinkedIn https://twitter.com/i/web/status/2019042438032306653 https://twitter.com/i/web/status/2019042438032306653"
X Link 2026-02-04T13:36Z 16.4K followers, 36.4K engagements

"Your API needs to return a user's transaction history with running balances but they have 1M transactions. How will you calculate and return this data to the client efficiently"
X Link 2026-02-04T18:26Z 16.5K followers, 20.4K engagements

"Your cache hit rate is 99% but your API is still slow. Users complain about 2-second response times. Where is the bottleneck and how will you fix it Real debugging scenario at Amazon"
X Link 2026-02-05T11:41Z 17K followers, 83.4K engagements

"Circuit Breaker Pattern in Distributed Systems Your Order Service calls Payment Service. Payment Service goes down. Without circuit breaker: Every order request waits [--] seconds for timeout All threads blocked waiting Order Service crashes from resource exhaustion Cascading failure kills your entire system With circuit breaker: Detects Payment Service is down after [--] failures Stops calling it immediately (fails fast) Order Service stays alive System recovers gracefully --- What Is a Circuit Breaker Think of your home's electrical circuit breaker. When there's a power surge the breaker "trips""
X Link 2026-02-05T12:58Z 15.9K followers, 18.9K engagements

"@ConsciousRide One can use a redis cache for rate limiting users with token bucket algorithm. Would need distributed redis for handling so much traffic (redlock algorithm comes to mind) Intersting read: https://redis.io/docs/latest/develop/clients/patterns/distributed-locks/ https://redis.io/docs/latest/develop/clients/patterns/distributed-locks/"
X Link 2026-02-05T13:14Z 17K followers, [---] engagements

"You cached user sessions in Redis but after a Redis restart [--] million users got logged out and your servers crashed from the login storm. How will you prevent this"
X Link 2026-02-05T16:48Z 17K followers, 82.3K engagements

"You need to rate limit your API to [---] requests per minute per user but you have [--] servers behind a load balancer. How will you enforce this limit consistently Asked at Stripe interview https://twitter.com/i/web/status/2019705106565062716 https://twitter.com/i/web/status/2019705106565062716"
X Link 2026-02-06T09:29Z 17.1K followers, 89.1K engagements

"This is easy. Your users table has 100M rows and SELECT * FROM users WHERE email = 'email protected' takes [--] seconds. How will you fix this"
X Link 2026-02-06T15:55Z 16.8K followers, 101.7K engagements

"FAANG system design at work doesnt look like interview system design. Interview: design Twitter in [--] min. Real life: keep [--] services alive while traffic shifts 3x one dependency is flaky and your DB is at 85% CPU. [--]. Constraints are real: quotas compliance multi-region migration windows org boundaries. [--]. Problems are boring but sharp: p99 from 900ms to 150ms hot partitions cache stampedes retry storms. [--]. Success is measurable: error budget SLOs cost per request oncall pages/week. [--]. You ship in slices: flags dark launches backfills rollback plans dashboards first. And yes you get paid for"
X Link 2026-02-06T20:41Z 17.3K followers, 17.6K engagements

"You added an index on email but the query is STILL slow. You run EXPLAIN and see "Seq Scan" instead of "Index Scan". What's wrong Production issue at Twitter now X This is easy. Your users table has 100M rows and SELECT * FROM users WHERE email = 'email protected' takes [--] seconds. How will you fix this This is easy. Your users table has 100M rows and SELECT * FROM users WHERE email = 'email protected' takes [--] seconds. How will you fix this"
X Link 2026-02-07T02:37Z 16.3K followers, 25.4K engagements

"You have a posts table with a view_count column that updates every time someone views a post. Popular posts get 1M views per day and your database is at 100% CPU. How will you reduce database load This problem has interesting solutions"
X Link 2026-02-07T08:13Z 17K followers, 26.2K engagements

"@RahulBagal__ That's correct what kind of redis instance/cluster would you use though"
X Link 2026-02-07T08:33Z 16.9K followers, [----] engagements

"@devXritesh That's correct what kind of redis instance/cluster would you use though"
X Link 2026-02-07T08:33Z 17.1K followers, [----] engagements

"Your payment API sometimes gets called twice due to network issues charging customers double. How will you make it idempotent Discuss exact steps Real bug in production at Shopify https://twitter.com/i/web/status/2020092094602531147 https://twitter.com/i/web/status/2020092094602531147"
X Link 2026-02-07T11:07Z 17.1K followers, 77.8K engagements

"@devXritesh Hmm I would rather generate the key on the backend side instead of the client side. And keep it in the database against the order ID because keeping it in Redis can cause problems as this is critical; we need a single source of truth"
X Link 2026-02-07T11:16Z 17K followers, [----] engagements

"@akashbitm787 Saving in DynamoDB would be a good option here. @grok can you discuss the pros and cons of using Redis DynamoDB or PostgreSQL here Since this is a critical thing I think PostgreSQL might be the go-to choice. Also consider generating it in the backend for each order"
X Link 2026-02-07T11:20Z 16.7K followers, [----] engagements

"Your API returns different responses for the same request: sometimes [---] with data sometimes [---] sometimes [---]. Users are confused and retrying makes it worse. What's the bug"
X Link 2026-02-07T14:15Z 17.1K followers, 54.4K engagements

"@ConsciousRide Yes we need proper tracing (use datadog)"
X Link 2026-02-07T14:37Z 17.3K followers, [---] engagements

"USA vs India salary growth (typical product companies) same ladder: [--]. SDE1: $120160k [----] LPA [--]. SDE2: $170240k [----] LPA [--]. SDE3: $250350k [----] LPA [--]. Staff: $350500k [--] LPA1.3 Cr [--]. Principal: $500900k+ 1.22.5 Cr Same titles. Different ceiling. So in India you often need "scope + leverage + equity" to match what US comp gives by default. Don't let the internet fool you about PPP and what not you definitely save much more in USA. And I won't even get started on the work culture. So go there if you get a chance and work for sometime. https://twitter.com/i/web/status/2020171828820672834"
X Link 2026-02-07T16:24Z 17K followers, 91.2K engagements

"Imho Caching is easy. Invalidation is where good systems falter. Some tips: [--]. Pick the right cache: browser CDN Redis in-process. Know what youre optimizing (p95 DB load cost). [--]. Use TTLs as a safety net not a strategy. If you need 24h TTL you will probably need better keys. [--]. Prefer versioned keys: user:123:v7 beats delete user:123 in distributed setups. [--]. Invalidate by event: write path emits UserUpdated(123) consumers bump versions or delete sets. [--]. For correctness serve stale with background refresh (stale-while-revalidate) instead of blocking hot paths. [--]. Guard against stampedes:"
X Link 2026-02-07T16:29Z 17.1K followers, [----] engagements

"Your API has proper authentication but users are able to access other users' data by changing the user_id in the URL. What's the vulnerability and how do you fix it OWASP Top [--] - Broken Access Control https://twitter.com/i/web/status/2020396775203565985 https://twitter.com/i/web/status/2020396775203565985"
X Link 2026-02-08T07:18Z 17.1K followers, 65.6K engagements

"@lalitkhu Yes one should test these extensively before pushing them to prod"
X Link 2026-02-08T08:04Z 16.1K followers, [----] engagements

"@ConsciousRide Correct test these endpoints fr"
X Link 2026-02-08T08:12Z 16.2K followers, [----] engagements

"@Umesh__digital Correct AWS secret manager is ok. But one should just keep them hashed"
X Link 2026-02-08T11:20Z 16.2K followers, [----] engagements

"Working to realize your potential as a software engineer is hard. But what is even harder is hitting [--] and realizing you had the talent the internet the time and you still chose comfort. No amount of take a break bro fixes the pain of knowing you could have built real leverage: - strong fundamentals - a portfolio that proves you can ship - a niche (distributed systems infra security data web3 whatever) - savings so you can take risks - a personal brand so opportunities find you Most engineers do not fail because they are dumb. They fail because they keep postponing the hard season. You do"
X Link 2026-02-08T14:18Z 17K followers, 72.1K engagements

"You have a chat application with WebSocket connections but when you deploy all 100K connected users get disconnected. How will you deploy without disrupting users Welcome to Monday https://twitter.com/i/web/status/2020721643958792593 https://twitter.com/i/web/status/2020721643958792593"
X Link 2026-02-09T04:49Z 17.1K followers, 59.4K engagements

"@ConsciousRide Redis state storage is important we don't want to lose the websocket connection information. @grok explain this please"
X Link 2026-02-09T05:23Z 16.6K followers, [---] engagements

"One should always first profile what exact command/value size is burning CPU then remove Redis from the 1M QPS read path via short TTL in-process caching or reduce payload. If it must stay in Redis duplicate the hot key across multiple instances and load-balance reads. One Redis key (product:123) is accessed 1M times/sec. Redis CPU hits 100%. What's your next step One Redis key (product:123) is accessed 1M times/sec. Redis CPU hits 100%. What's your next step"
X Link 2026-02-09T07:41Z 16.7K followers, [----] engagements

"@devXritesh Small wins compound. Keep going"
X Link 2026-02-09T08:13Z 16.6K followers, [---] engagements

"@Doom_S_Dey well explained @grok can you explain blue green deployment for the uninitiated"
X Link 2026-02-09T09:24Z 16.6K followers, [----] engagements

"Your real-time notifications work fine with [---] users but with 10K users messages take [--] seconds to arrive. What's the bottleneck Asked at Slack interview https://twitter.com/i/web/status/2020797825731440767 https://twitter.com/i/web/status/2020797825731440767"
X Link 2026-02-09T09:52Z 17.1K followers, 45.5K engagements

"Super Bowl traffic is not just high traffic. Its a spike with money attached. We started hitting rate limits mid-game because: [--]. Every bet triggered [--] to [--] backend calls. [--]. Fraud checks fanned out to multiple vendors. [--]. Retries turned a small slowdown into a thundering herd. Fix in the moment: [--]. Raised rate limits (carefully per route for the specific betting platform customer). [--]. Scaled instances 3x. [--]. Added a temporary queue for fraud so bets didnt block on vendor latency. Peak events break your assumptions. Load test the fan-out paths not just the bet. #SuperBowlAPI #SuperBowlLX"
X Link 2026-02-09T10:03Z 16.8K followers, [----] engagements

"@YoelNisanov Exactly"
X Link 2026-02-09T11:10Z 16.9K followers, 10.4K engagements

"You have a distributed counter (total views on a post) across [--] servers. Each server increments locally and syncs to the database every [--] seconds but the total is always wrong. How will you fix this Classic distributed systems problem"
X Link 2026-02-09T13:37Z 17.1K followers, 14.3K engagements

"@devXritesh Redis incr 🔥"
X Link 2026-02-09T13:43Z 17K followers, [----] engagements

"@akashbitm787 💯 correct. Atomic update with redis incr or update query shared above should work"
X Link 2026-02-09T13:45Z 17K followers, [---] engagements

"Your microservices use UUID primary keys. When you try to join data from two services queries are 100x slower than with integer IDs. What's happening and how do you fix it"
X Link 2026-02-09T17:00Z 17.1K followers, 138.5K engagements

"You deployed a new feature that adds a row to the database but 10% of rows are missing in production even though logs show successful inserts. What could cause this Real production mystery at Netflix"
X Link 2026-02-09T17:03Z 17K followers, 31.3K engagements

"@TreeApostle Nice. Yes this is the reason here. Use both. Big ints for joins and uuids for other purposes. We recently implemented a postgres equivalent for storing clickhouse audit logs and used both for fast insertions"
X Link 2026-02-09T17:18Z 17K followers, [----] engagements

"@brankopetric00 Dump it all into ClickHouse and nothing can hurt you later on"
X Link 2026-02-09T20:15Z 17K followers, [---] engagements

"I connected with someone on LinkedIn (can't take the name for privacy purposes) and they asked me to review their resume. Here's how it was structured. --- Led a backend API-driven data ingestion project enabling historical data uploads reducing customer go-live time by 70% and improving onboarding scalability Automated customer onboarding workflows by developing Jenkins jobs based on Python and shell scripts reducing on-boarding time by 50% Migrated 50+ co-sell with hyperscaler customers and simplified package configurations for smooth experience Implemented unit testing for AWS Lambda"
X Link 2026-02-10T03:54Z 17K followers, 21.4K engagements

"Your background job queue has 1M jobs but workers are only processing [--] jobs per minute. How will you speed this up"
X Link 2026-02-10T06:48Z 16.9K followers, 40.1K engagements

"@phillipsharring Yikes"
X Link 2026-02-10T06:55Z 16.8K followers, [---] engagements

"Those who answered to use UUIDv7 are wrong. They didn't think about the migration path. UUIDv7 does not solve our existing tables pain. UUIDv7 helps mostly going forward because the big win is insert order and index locality. But Our existing table built on random UUIDs is already fragmented and wide. Switching generators does not magically fix that. To truly fix old data we would still need a heavy table rewrite / re-cluster (and downtime risk) which is basically the migration we are trying to avoid. So if we have large existing tables and we want the fastest lowest risk fix: - Add BIGINT"
X Link 2026-02-10T07:08Z 16.9K followers, 11.3K engagements

"@Umesh__digital Correct identify bottleneck first add metrics see datadog create dashboards add alerts look for mem or cpu spike"
X Link 2026-02-10T07:17Z 16.8K followers, [---] engagements

"@omranchi00 Haha 🌎 Please debug this on planet Earth only"
X Link 2026-02-10T07:19Z 16.8K followers, [---] engagements

"@code_benji Diagnosis is always the first step"
X Link 2026-02-10T08:44Z 16.8K followers, [--] engagements

"Your Kafka consumer is processing messages out of order causing data corruption. How did this happen and how do you ensure ordering Production issue at Uber"
X Link 2026-02-10T10:14Z 17K followers, 44.2K engagements

"@codewiddan That's right. @grok can we use sqs does it guarantee ordering"
X Link 2026-02-10T10:28Z 16.8K followers, [----] engagements

"Game theory [---]. Especially if you missed those classes during your semester (it was clubbed with pretty useless subjects like environmental science in my college). Bitcoin miners are like factories. They spend real money (electricity rent machines staff) to produce Bitcoin rewards. Their income is paid in Bitcoin. So if Bitcoin price drops a lot their revenue in dollars drops instantly but most of their bills stay the same. Same Bitcoin miners are like factories. They spend real money (electricity rent machines staff) to produce Bitcoin rewards. Their income is paid in Bitcoin. So if Bitcoin"
X Link 2026-02-10T10:37Z 16.9K followers, [----] engagements

"@gujjutweeter Haha that's correct theoretically but not scalable"
X Link 2026-02-10T11:00Z 16.8K followers, [---] engagements

"@patilvishi That's right we can keep a single partition but that's not scalable. @grok can you state the latest version of kafka that has support for message ordering too"
X Link 2026-02-10T11:01Z 16.8K followers, [---] engagements

"You're using RabbitMQ and a message was processed successfully but the server crashed before acknowledging. Now the message is being processed again. How do you handle this Classic message queue problem"
X Link 2026-02-10T12:51Z 17K followers, 47.7K engagements

"@RahulBagal__ That's correct. Using the outbox pattern can be beneficial here but idempotency is more important even if you don't have the outbox pattern implemented"
X Link 2026-02-10T12:57Z 16.8K followers, [----] engagements

"@func25 Testcontainers in Go. But I need to check how we can use it in our CI/CD pipeline. Maybe @grok can help here"
X Link 2026-02-10T13:38Z 16.9K followers, [----] engagements

"@devXritesh That's true. Apparently that guy found another job at goldman sachs"
X Link 2026-02-10T17:01Z 16.9K followers, [----] engagements

"@rajat_namespace For multiple retries. @grok can you take this up"
X Link 2026-02-11T06:31Z 16.9K followers, [---] engagements

"I needed one compute box for dev. So I spun up a single EC2 instance. SSH in run the app. Simple. Then we went to production. Just [---] users but uptime matters. So I put the app behind an ALB and ran [--] instances in an Auto Scaling Group across [--] AZs. If one dies it heals itself. Traffic grew. CPU spikes latency up. So I scaled out (2 to [--] instances) with autoscaling rules. Same code more capacity. Deploys got painful. SSH deploys are a trap. So I containerized pushed to ECR and deployed by image tag. Rollbacks became easy. Then we added more services and it got messy managing ports and"
X Link 2026-02-11T09:30Z 17K followers, [----] engagements

"Your service is failing health checks and getting killed by Kubernetes but when you check manually it's responding fine. What's wrong Common Kubernetes debugging scenario"
X Link 2026-02-11T10:29Z 17K followers, 10.2K engagements

"@ConsciousRide yes exactly imo its best to work this way: remote work work in USA work in india work in EU"
X Link 2026-02-11T11:22Z 17K followers, [---] engagements

"We also hired a guy around the same time who was genuinely strong technically. [--]. Wrote very good painful-to-break tests. [--]. Caught edge cases in reviews that saved us outages. [--]. Clean commits boring in the best way. But the role was 70% communication. Design docs stakeholder syncs driving alignment writing updates that unblock [--] people. He stayed quiet skipped context didnt push back early. Work was good but the team was always surprised. We fired him. [--] months later he joined CRED and hit SDE3 fast. In that setup his strengths printed money. Mismatch hurts both sides. Join where your"
X Link 2026-02-11T11:33Z 17K followers, 21.9K engagements

"This problem came up during our weekly architecture call. We kept scaling our notification consumers by adding more pods and every pod held its own Kafka connection. The end result was tons of pods tons of connections and mediocre utilization. Later we realized this workload is mostly IO-bound not CPU-bound. So we flipped it: fewer pods and inside each pod we increased concurrency with more goroutines (and sane limits). Same throughput lower resource waste better CPU/mem usage simpler ops. So before you horizontally scale check if you just need more in-process concurrency. In case of golang"
X Link 2026-02-11T11:51Z 17K followers, [----] engagements

"Tbh people taking up this course are probably dumb. No offense to @gkcs_ I have followed him from his early days. And definitely no offense to @striver_79 for promoting a course that is not making fake promises because he himself is in the course selling space. But if you are taking a course to "learn AI" which is very much self-learnable (i.e. we can use AI to learn AI) then you can never get better at learning things yourself. In my opinion one should take these fast-changing times as a signal to get better at learning things. And anybody with common sense can learn to use AI to automate"
X Link 2026-02-11T13:50Z 17K followers, 18.1K engagements

"Users report that sometimes the app is fast (100ms) and sometimes it's very slow (5+ seconds) but you can't reproduce it. How do you debug this Asked at Google interview"
X Link 2026-02-11T14:27Z 17K followers, 55.7K engagements

"Unpopular opinion: There is no such thing as AI agentic engineering; it's normal good old engineering powered by AI. Anyone who tells you they are an AI engineer now is just a celebrated prompter. Period. If you want to learn math behind ML that's something which will make you a deep learning engineer for real. Follow @karpathy to learn that. Instead of using terms like "agnetic AI" and glorifying LLM token generation with models that were actually developed by true AI pioneers it might be more appropriate to simply refer to this technology as "LLM." Tbh the AI Engineer title is just a cheap"
X Link 2026-02-11T18:29Z 17K followers, [----] engagements

"@akashbitm787 Imo that's just adding another layer of complexity to handle something that's already working fine in prod deterministically"
X Link 2026-02-11T20:08Z 17K followers, [---] engagements

"@GautamK0508 Nice buying asap"
X Link 2026-02-11T20:33Z 17K followers, [--] engagements

"You need to send an email when a user signs up but the email service is slow (5 seconds per email) and blocking user registration. How will you fix this Basic async processing question"
X Link 2026-02-11T21:52Z 17.1K followers, 16.9K engagements

"@ConsciousRide I will let a FE guy handle it (like yourself) and use analytics db like clickhouse or redshift in the backend to return aggregated data points"
X Link 2026-02-11T23:37Z 17K followers, [---] engagements

"That something big is just market cycles coming to an end. Welcome to the great financial reset. Last time the narrative was covid now it's AI. We will see a big crash followed by the start of a new cycle. AI hype is not real. As a developer i can say for sure. We need managers and who will be the manager. The Devs will step up to manage things end to end. So become a better manager. Forget the rest. There are always bigger forces at play. Learn to see the pattern now. https://twitter.com/i/web/status/2021733051060076993 https://twitter.com/i/web/status/2021733051060076993"
X Link 2026-02-11T23:48Z 17K followers, [----] engagements

"@RahulBagal__ It's not important to become staff - it's your choice - but if you do want to aim for it you need to be consistent with the points listed above"
X Link 2026-02-12T08:17Z 17K followers, [---] engagements

"@omranchi00 "Staff is a mindset" - well put"
X Link 2026-02-12T08:18Z 17K followers, [--] engagements

"@omranchi00 That's correct. Or you can keep buying course after course - ngmi"
X Link 2026-02-12T08:19Z 17K followers, [--] engagements

"Some people keep saying that "agentic AI" (or "AI engineering") basically means running production systems deterministically - like were supposed to wrap a chaotic black box in enough guardrails until it behaves like a normal service. But when did we even get to that point Things were working fine before. We already know how to build reliable systems: clear specs deterministic code observability good tests tight feedback loops. So why are we voluntarily adding a new layer of complexity that we dont fully understand then pretending its just another dependency Im all for using AI to boost"
X Link 2026-02-12T08:46Z 17.1K followers, [----] engagements

"Your application needs to process uploaded images: resize compress generate thumbnails. This takes [--] seconds and users wait. How will you improve user experience"
X Link 2026-02-12T09:39Z 17.1K followers, 11K engagements

"1. Same salary. [--]. Same location. [--]. Same codebase problems. [--]. Same manager conversations. [--]. Same next quarter promises. [--]. Same hike that disappears in rent and groceries. Stability is fine. But if nothing changes for [--] to [--] months you are not being consistent. You are being priced in. change scope change leverage or change company"
X Link 2026-02-12T10:13Z 17.1K followers, [----] engagements

"You have a monolithic application that takes [--] minutes to deploy and any bug takes down the entire system. How would you begin breaking this into microservices System design interview classic Thanks for participating - this is the last question of the series will come up with a new set of quality questions from next week"
X Link 2026-02-12T13:32Z 17.1K followers, [----] engagements

"@ConsciousRide I would use Redis to just store the blocked users by an account. When showing the feed I would just check against the array. Because one account cannot block 1000s of accounts but only a few 100s"
X Link 2026-02-12T13:53Z 17K followers, [---] engagements

"Now this is AI engineering No paid course can teach this stuff. New art project. Train and inference GPT in [---] lines of pure dependency-free Python. This is the full algorithmic content of what is needed. Everything else is just for efficiency. I cannot simplify this any further. https://t.co/HmiRrQugnP New art project. Train and inference GPT in [---] lines of pure dependency-free Python. This is the full algorithmic content of what is needed. Everything else is just for efficiency. I cannot simplify this any further. https://t.co/HmiRrQugnP"
X Link 2026-02-12T17:10Z 17.1K followers, 37K engagements

"Countries With the Largest Software Developer Populations [--] China - 4.39M [--] United States - 2.90M [--] India - 1.92M [--] Japan - 1.27M [--] Brazil - 0.84M [--] Germany - 0.56M People confuse this with tech workers. This is closer to professional software developers. It excludes a lot of adjacent roles (IT ops support product design analysts) and it also misses hobby devs. Also these numbers are always estimates. Different reports count different things (full-time only vs contractors self-employed students etc). Still directionally it shows where the biggest engineering supply pools are forming."
X Link 2026-02-12T18:00Z 17.1K followers, 12.1K engagements

"RT @karpathy: Congrats on the launch @simile_ai (and I am excited to be involved as a small angel.) Simile is working on a really intere"
X Link 2026-02-12T20:16Z 17K followers, [---] engagements

"@ConsciousRide Imho this is just a narrative we are closer to recession than most even realise and this is just the beginning instead of focusing on such narratives (AI aliens etc.) focus on building stuff and honing your fundamentals"
X Link 2026-02-13T09:59Z 17.1K followers, [---] engagements

"@devXritesh Slack one is very good. Thanks for adding these"
X Link 2026-02-13T12:26Z 17.1K followers, [---] engagements

"I usually ask: What is the biggest problem on your plate right now that is still not solved You see this reveals the real pain. Not the marketing version of the job listed on the careers page. Asking questions is non-negotiable: - It shows you are evaluating them too not begging for a job - It gives you signal on ownership culture and how real the role actually is The interview is not just them testing you. It is you doing due diligence on where you will spend your next 1-2 years. Whenever the interviewer asks "Do you have any questions for me" never skip this question. Here are [--] questions"
X Link 2026-02-13T12:54Z 17.1K followers, [----] engagements

"I feel like [----] is getting priced in. Recession talk is back AI narrative is still doing the "every company is an AI company" thing and the recent Google buyout news is another reminder that Big Tech is optimizing for cashflow. If youre a software dev act like the market is tighter for [--] to [--] months: [--]. Get painfully good at one business-critical lane (payments infra reliability data pipelines security). Generalist is fine but owns a metric wins. [--]. Ship things that save or make money. Track numbers: p95 down 30% cloud bill down 15% fraud loss down 5%. Put it in your brag doc. [--]. Treat AI"
X Link 2026-02-13T12:57Z 17.1K followers, 12.6K engagements

"Interviewer: Design a Google-scale system in [--] minutes. Me: Respectfully if you gave me ten years I still wouldn't be able to do that"
X Link 2026-02-13T14:44Z 17.2K followers, 22.3K engagements

"@Manan_13 All the best"
X Link 2026-02-13T18:10Z 17.2K followers, [--] engagements

"Quick notes on event-driven architecture with Kafka. [--]. Producers write events to a topic (append-only log). [--]. Partitions give you scale but also ordering only inside a partition. [--]. Consumers read with a group id. Kafka load-balances partitions across the group. [--]. Offsets are your checkpoint. Commit is done only after work is done. [--]. Expect duplicates. Make handlers idempotent. Use a key dedupe table or upsert. [--]. Add retries + DLQ for poison messages. Set timeouts. Watch consumer lag. Kafka gives throughput. But you should handle correctness Part of my quick notes series"
X Link 2026-02-13T18:20Z 17.1K followers, [----] engagements

"@jmtame lol even claude would stumble here haha"
X Link 2026-02-13T18:26Z 17K followers, [---] engagements

"@kirat_tw If you have a low iq and don't have any physical brain problems. You can get a higher iq. Ergo anyone can learn anything"
X Link 2026-02-15T08:45Z 17.2K followers, [---] engagements

"My playbook for going from junior engineer to senior engineer: [--]. Master the fundamentals (data structures networking basics databases OS concepts). Weak foundations cap growth fast. [--]. Get very good at one primary language and ecosystem used at work. Depth beats shallow polyglotting early on. [--]. Take ownership of small features end-to-end: design implementation tests deployment and monitoring. [--]. Read and understand the existing codebase deeply before trying to improve it. Context matters. [--]. Ask why things are built the way they are. Trade-offs are where senior thinking starts. [--]. Learn to"
X Link 2025-12-21T19:20Z 17.3K followers, 83.4K engagements

"@jungleeBilla6 Keep it a single page. Short and succinct"
X Link 2026-02-10T04:55Z 17.3K followers, [---] engagements

"Using AI to learn AI - wagmi Using Courses to learn AI - ngmi https://twitter.com/i/web/status/2021859791556198469 https://twitter.com/i/web/status/2021859791556198469"
X Link 2026-02-12T08:11Z 17.3K followers, [----] engagements

"We had a Go service doing NFS S3 uploads. [--]. Used io.ReadAll() to read the whole file into memory. [--]. Worked fine in dev crashed in prod. [--]. [---] MB files + a few concurrent uploads = GBs of RAM GC thrash random OOMs. Fix was boring: [--]. Stop buffering. [--]. Stream: NFS reader S3 uploader (io.Copy / multipart). [--]. Constant memory stable p99 no surprises. My learning: when working with files assume streams until proven otherwise We had an S3 upload path in Go doing io.ReadAll() first. [--]. It doubled memory. One big byte then another copy when building the request. [--]. Under concurrency RSS spiked"
X Link 2026-02-12T13:37Z 17.3K followers, 27.1K engagements

"@motorcycle_guy Pro tip"
X Link 2026-02-14T15:02Z 17.3K followers, [---] engagements

"Sagas are useful when a business flow spans multiple services and failures are normal so you need compensations. But if you control the boundary it is often better to keep the whole flow in one service + one DB and publish events outward. Less coordination less partial failure hell. https://twitter.com/i/web/status/2022726499095494727 https://twitter.com/i/web/status/2022726499095494727"
X Link 2026-02-14T17:35Z 17.3K followers, [---] engagements

"@haha_girrrl My connection pools are available if you want to connect after restarting the server"
X Link 2026-02-14T19:54Z 17.3K followers, [----] engagements

"Thanks for all the nice discussions my answer. Use sticky sessions (route user to same server via cookie). But better: centralize session storage in Redis/Memcached (all servers share session data). Use session token in cookie pointing to Redis key. For scale: Redis cluster with replication. Alternative: client-side sessions (JWT with encrypted cart data). Trade-off: Sticky sessions easier but less resilient (server restart = lost carts). Centralized storage more complex but robust. Add session replication for critical data. https://twitter.com/i/web/status/2022962735705956779"
X Link 2026-02-15T09:14Z 17.3K followers, [---] engagements

"@vibeonX69 I am a nomad will work from Chandigarh for a week then shimla then go back to my base in bglr and chill out with my flatmate @akashbitm787"
X Link 2026-02-15T10:15Z 17.3K followers, [---] engagements

"@d4rsh_tw Lol I thought cf specialists can at least be interns at Amazon"
X Link 2026-02-15T17:29Z 17.3K followers, [---] engagements

"@VazeKshitij It's sad. Not that he wasn't on Amazon but because people followed him just on the basis of a tag. I followed him because of 1-2 good tweets and because he was verified. We need a credibility layer asap"
X Link 2026-02-15T17:31Z 17.3K followers, [---] engagements

"@Pragyan2214 Thanks pragyan"
X Link 2026-02-16T11:28Z 17.3K followers, [--] engagements

"The common approach is wait-for graph analysis: Every thread is a node every lock is a resource. If thread A is waiting on a lock held by thread B you add an edge A B. If that graph develops a cycle youve got a deadlock. This is more precise than just timing out timeout-based detection can give false positives while graph-based cycle detection actually proves that progress is impossible. Some DBs and OS kernels combine both: cycle detection for correctness timeouts for responsiveness"
X Link 2025-09-06T12:23Z 17.3K followers, [---] engagements

"Same job. [--] companies. [--] different titles. [--]. Infosys (Bangalore) Title: Systems Engineer Work: bug fixes support tickets small features Salary: [---] LPA [--]. Amazon (Hyderabad) Title: SDE I Work: owned [--] service on-call wrote design docs shipped weekly Salary: [--] LPA (base + bonus) [--]. Razorpay (Bangalore) Title: Backend Engineer Work: payments workflows idempotency retries incident rotations Salary: [--] LPA (CTC) [--]. Uber (Bangalore) Title: Software Engineer II Work: scaled APIs reduced p99 deep debugging cross-team projects Salary: [--] LPA (base + RSUs) [--]. US startup (Remote India) Title: Senior"
X Link 2026-02-12T12:56Z 17.3K followers, 288.6K engagements

"I have seen most software devs can do derivatives trading some of them just copy random strategies and blow up. Imho it should be treated like engineering: [--]. Pick [--] market (NIFTY options SPX BTC perps). Learn the contract specs margin settlement. [--]. Start with risk code: max loss per day (1R) position sizing stop rules no averaging. [--]. Automate journaling first not execution. Log entry thesis IV greeks PnL slippage time held. [--]. Build a small edge: vol regime filters simple spreads event-driven rules. Backtest honestly include fees. [--]. Paper trade [--] days then go tiny. Survive [--] months"
X Link 2026-02-14T12:20Z 17.3K followers, [----] engagements

"Append every reaction/comment as an event into Kafka/Pulsar with partitions by match_id and time-bucket so you can sustain 50M concurrent without hot partitions. Raw events should land in an OLAP store (ClickHouse) for rollups + audits and a fast KV (Redis/KeyDB) for per-second counters (emoji tallies poll votes). For fan-out: one should never push every comment to every user we publish only streams (top comments friends local language verified stadium cams) using a pub/sub layer (Redis Cluster / NATS / Kafka consumers) where each stream is pre-aggregated + ranked continuously. Clients"
X Link 2026-02-14T15:36Z 17.3K followers, [----] engagements

"25 programming languages and what theyre good for: [--]. C: kernels embedded I need control. [--]. C++: low-latency systems games trading. [--]. Rust: safe systems perf without footguns. [--]. Go: backend services networking boring scale. [--]. Java: big enterprises JVM ecosystems stable ops. [--]. Kotlin: Android modern JVM backend. [--]. C#: Windows Unity solid web APIs. [--]. Python: automation data fast prototypes. [--]. JavaScript: browser fullstack glue code. [--]. TypeScript: JS at scale fewer production bugs. [--]. PHP: web apps cheap hosting fast CRUD. [--]. Ruby: Rails product velocity startups. [--]. Swift: iOS"
X Link 2026-02-14T15:59Z 17.3K followers, 41.8K engagements

"If youre a remote software engineer you need [--] things: Reliable WiFi sane rent and a life that doesnt drain you. [--]. Tbilisi Georgia $600$1000/month Great value solid cafs easy long stays dev-friendly vibe. [--]. Da Nang Vietnam $500$900/month Beach city + low cost + good internet. Quiet enough to actually ship. [--]. Chiang Mai Thailand $700$1200/month The classic builder city. Cheap comfy huge remote community. [--]. Kuala Lumpur Malaysia $800$1400/month Modern city great food strong infrastructure underrated for devs. [--]. Lisbon (or Porto) Portugal $1300$2200/month More expensive but high quality of"
X Link 2026-02-14T17:45Z 17.3K followers, 36.6K engagements

"Zoom out and look at your progress as a software engineer on a longer timescale. Day by day it feels like nothing is changing. Same tickets bugs same standups same feeling that you are still behind. But compounding is quiet. You do not notice it while it is happening. Then one day you look back [--] months [--] year [--] years and you realize you are not the same engineer anymore. You read logs faster. You debug faster. You ask better questions. You ship with less fear. You understand trade offs. You see the system not just the function. I have a friend who used to feel stuck like this. He was a"
X Link 2026-02-15T14:15Z 17.3K followers, [----] engagements

"X goes down for [--] minutes and you can feel it. [--]. Infra team please triage. [--]. Which infra team [--]. The one that built it in [----]. [--]. Yeah they left. I feel that X lost the people who knew where those weird cron job and even more complex playbooks live"
X Link 2026-02-16T14:32Z 17.3K followers, [----] engagements

"@TrumpTruthOnX Heartbreaking. Whatever your politics this is a tragedy. A husband father friend is gone. We need less heat more humanity. RIP Charlie Kirk. Prayers for Erika and the family"
X Link 2025-09-10T20:58Z 17.3K followers, 12.6K engagements

"$4.21B expiring automatic pump or dump. Options expiry is about flows: if dealers are long gamma they hedge against you spot gets pinned near big strikes (max pain vibes) if dealers are short gamma hedging chases price moves can amplify after a break Practical read for today: Watch where the biggest strikes sit (the magnets) and how price behaves into them Check perp funding basis and spot vs OIif OI unwinds and spot leads move is cleaner After expiry hedges come off air pockets appear post-expiry move often matters more than the [--] min headline Big lesson for new traders: size of notional"
X Link 2025-09-12T08:32Z 17.3K followers, [----] engagements

"My superpower as a remote software dev The real superpower of being a remote developer isnt just freedom its the optionality it gives you. I have been working from landour for a week now and realised how blessed we are to have this kind of profession. You can work from anywhere and more importantly learn from anywhere. Ive spent mornings writing Go APIs from a small caf afternoons debugging infra issues with a view of the sea and nights going deep into Rust docs to create my youtube videos just because I could. Remote work lets you carve time for what truly compounds - niche skills. Things"
X Link 2025-10-16T05:48Z 17.3K followers, 131.6K engagements

"If your goal is backend mastery you dont need [--] projects. You need [--] projects that force you to think like a backend engineer. Here are the most important [--] in order [--]. Build your own Web Server HTTP parsing routing middleware concurrency timeouts. This teaches you how every backend request actually flows. [--]. Build your own Database / Key-Value Store Storage engine indexing WAL compaction concurrency control. Youll finally understand why databases are slow or fast. [--]. Build your own Cache System (Redis-like) Eviction policies (LRU/LFU) TTLs memory management persistence tradeoffs. This"
X Link 2026-01-08T13:31Z 17.3K followers, 120.2K engagements

"My friend moved to Seattle for Amazon. $170K total comp. On-call every [--] weeks. Pager at [--] AM. Customer obsessed on the wall burnout in the body. Then the layoff email came. [--] minutes with HR. Laptop was shipped back. Access revoked before the call ended. What surprised him: [--]. Performance didnt matter that day. Headcount math did. [--]. Your manager might fight for you but Finance wins. [--]. Your network is real only if you built it before you need it. [--]. Severance is a runway not a vacation. He started interviewing in [--] hours. He told me: I thought that there was job security. It was just a"
X Link 2026-01-28T20:21Z 17.3K followers, 4.5M engagements

"Your database has a users table and an orders table. You need to show "users who haven't placed an order in [--] days" but the query times out. How will you optimize this Asked at Airbnb interview https://twitter.com/i/web/status/2019964214241288697 https://twitter.com/i/web/status/2019964214241288697"
X Link 2026-02-07T02:39Z 17.3K followers, 74.4K engagements

"Explaining these with top level points and small examples. [--]. Event-Driven Architecture Top points: publish/subscribe async loose coupling retries + duplicates happen. Mini example: OrderService emits OrderPlaced to a broker. InventoryService consumes it and reserves stock. If InventoryService is down the event is retried later so your consumer must handle duplicates (same order might be processed twice) using an idempotency key like order_id. [--]. Saga Pattern Top points: distributed transaction without 2PC step-by-step workflow compensations orchestration vs choreography. Mini example: Place"
X Link 2026-02-07T10:33Z 17.3K followers, 19.3K engagements

"Your API keys are stored in the database as plain text. A database dump leaked and all API keys were compromised. How should API keys be stored As a software dev this is just "plain" funny https://twitter.com/i/web/status/2020453880438599945 https://twitter.com/i/web/status/2020453880438599945"
X Link 2026-02-08T11:05Z 17.3K followers, 145.6K engagements

"We hired a backend guy recently who didnt know half the buzzwords. No Saga no CQRS shaky on K8s. On paper easy reject. Then we gave him a real prod-ish bug: sporadic 500s p95 spikes only on one endpoint. He did [--] things: [--]. Asked for repro + timeline. When did it start What changed Any new feature release [--]. Cut the problem space. Logs first then metrics then a single failing request ID. [--]. Formed a hypothesis tested it wrote down what each result would mean. Found it in [--] mins: connection pool exhausted from one code path leaking retries + no timeout. Ill take that over memorized concepts"
X Link 2026-02-09T09:30Z 17.3K followers, 854K engagements

"Your application is running fine but suddenly the CPU spikes to 100% every day at [--] AM for exactly [--] minutes. What could this be and how do you investigate"
X Link 2026-02-11T06:42Z 17.3K followers, 277.6K engagements

"The most likely career trajectory you will follow as a software engineer is the one your habits are already drawing for you every single day. Not your motivation. Not your one heroic sprint. Your defaults. If your default habit is ship small things consistently you slowly become the person who can ship big things under pressure. If your default habit is learn deeply and write things down you slowly become the person others depend on. If your default habit is avoid hard problems you slowly become the person who stays stuck even with talent. Becoming Staff is all about habits upgraded. [--]. Staff"
X Link 2026-02-12T08:05Z 17.3K followers, 17.1K engagements

"We had an S3 upload path in Go doing io.ReadAll() first. [--]. It doubled memory. One big byte then another copy when building the request. [--]. Under concurrency RSS spiked fast and pods were restarting due to OOM issue. Fix was boring: [--]. Stop buffering. [--]. Stream straight to S3 using the io.Reader interface. Memory footprint dropped by 50% immediately. If the API accepts io.Reader use it that's the beauty of Go interfaces. Dont help it with io.ReadAll(). Plug the stream in"
X Link 2026-02-12T09:44Z 17.3K followers, 30.4K engagements

"Recently a friend of mine appeared for Uber interview for L5 level and couldn't make it. Uber interview loops are pretty standardized. Your outcome mostly depends on level mapping. Heres what Ive seen for L3 to L6 (India + US ranges). Not exact but close enough to sanity-check an offer. [--]. Typical interview loop (Backend) 1) Recruiter screen (level + comp expectations) 2) [--] min DSA (LeetCode medium [--] problem solid edge cases) 3) [--] min DSA or coding + debugging (often a twist: constraints scaling memory) 4) [--] min System Design (L4+). L5 expects tradeoffs failure modes rollout. 5) [--] to 60"
X Link 2026-02-12T13:13Z 17.3K followers, 76.3K engagements

"Crazy this is why. [--]. Spacecraft failures are permanent: no restart no hotfix no physical access. [--]. These rules reduce surprises and force deterministic predictable behavior. [--]. No recursion = no unbounded stack growth and easier worst-case reasoning. [--]. Bounded loops = guaranteed termination and provable timing deadlines. [--]. No malloc/free after init = no fragmentation no random allocation failures stable latency. [--]. Small functions = easier code review testing and static analysis coverage. [--]. Assertions = early crash on impossible states instead of silent corruption. [--]. Checking every"
X Link 2026-02-12T18:04Z 17.3K followers, 17.1K engagements

"Kafka had an 8MB message limit. We hit it with a few "harmless" payloads. So we added a producer fallback: [--]. Try publish normal event. [--]. If payload too big write blob to S3. [--]. Publish to Kafka: type id s3_uri sha256 size. [--]. Consumers fetch from S3 verify hash process then delete/TTL. Kafka is for pointers and intent. Put big bytes in object storage keep events small and reliable. Thank me later"
X Link 2026-02-13T07:07Z 17.3K followers, 12.6K engagements

"Design a URL shortener for 1B links 50K writes/sec 200K reads/sec. Now a celebrity drops a link and one code gets 5M clicks in [--] minutes. What breaks first Some things to take care of: [--]. Hot key: cache stampede single shard overload [--]. DB: write amplification secondary index pain replication lag [--]. App: connection pool exhaustion thread starvation [--]. Network: LB limits TLS CPU egress cost [--]. Correctness: idempotency on create rate limits abuse And if you jump to use "Cassandra" without walking the interviewer through failure modes youre not designing correctly. Talk bottlenecks + numbers +"
X Link 2026-02-13T09:51Z 17.3K followers, 13.7K engagements

"Most devs read engineering blogs like entertainment. If you want skill read like youre on-call. [--]. Dont binge it. Pick [--] post/week and extract the model: load pattern constraints tradeoffs failure modes. [--]. Try rewriting the architecture in [--] lines. If you cant you didnt understand it. [--]. Then turn it into a checklist for your system: timeouts retries idempotency backpressure observability. [--]. Replicate one thing in a toy repo in [----] mins. Even a stub teaches more than [--] posts. [--]. Keep a notes file with: problem decision metric impact (p95 error rate cost) what youd do differently. Blogs"
X Link 2026-02-13T12:13Z 17.3K followers, [----] engagements

"Building backend systems is something I really care about and find fun - thinking through edge cases and fixing production-level bugs. But for those who dont know me personally Ive also been a smart contract engineer for 3years (not doing now but upto last year i was). The only reason i picked up solidity was for earning a lot of money quickly and it was solidity gold rush 2-3yrs ago. Ive shipped in DeFi and spent way too much time reading protocols: [--]. Uniswap [--]. Aave [--]. Compound [--]. OlympusDAO (and other weird fun stuff) Ill be posting small code snippets and some protocol level breakdowns."
X Link 2026-02-13T18:41Z 17.3K followers, [----] engagements

"I asked my friend a Golang system design question in a Zomato Blinkit context: An order event gets published twice. Driver app retries for [--] minutes. Inventory service is down for [--]. How do you stop double charging and still deliver He didnt talk about Go syntax or Goroutines or any Go-shenanigans These were his points. [--]. Idempotency keys on PlaceOrder and ChargePayment (stored in Postgres/Redis with TTL) [--]. Outbox pattern in the Orders DB then Kafka publish. No write to DB + publish race [--]. Consumer dedupe using (order_id event_id) and exactly-once illusion with at-least-once + idempotent"
X Link 2026-02-14T12:22Z 17.3K followers, 15.5K engagements

"I see many fresh software engineers panicking about AI. They think: If the model can write code what is my value But AI did not delete software engineering. It deleted low-trust engineering. In the AI era the currency is: can people trust you to ship outcomes. Heres how a beginner engineer becomes valuable (and hard to replace): [--]. Become a problem clarifier not a code typist Most juniors jump straight to implementation. The real skill is turning vague pain into a crisp spec: inputs outputs constraints edge cases success metrics. [--]. Own one thin slice end-to-end Pick a small feature and do"
X Link 2026-02-14T15:31Z 17.3K followers, 24.6K engagements

"25 databases and what theyre good for (picked from stuff Ive actually seen in prod): [--]. Postgres: default OLTP constraints SQL you can trust [--]. MySQL: OLTP at scale lots of tooling boring in a good way [--]. SQLite: local apps mobile tests one file DB [--]. SQL Server: enterprise OLTP + reporting in Microsoft shops [--]. Oracle: big regulated orgs expensive but battle-tested [--]. MariaDB: MySQL-compatible with different ops tradeoffs [--]. CockroachDB: distributed SQL multi-region strong consistency [--]. YugabyteDB: Postgres-ish distributed SQL HA + geo [--]. TiDB: MySQL-ish distributed SQL horizontal scale 10."
X Link 2026-02-14T16:04Z 17.3K followers, 20.9K engagements

"@Axel87830417 All of them are good enough @grok can you answer this"
X Link 2026-02-14T16:12Z 17.3K followers, [---] engagements

"@iamgoroot option2 is much better"
X Link 2026-02-14T16:18Z 17.3K followers, [---] engagements

"Your API Gateway becomes a bottleneck at 100K requests/sec. Adding more gateway instances doesn't help because they all hit the same rate-limiting database. How will you scale rate limiting"
X Link 2026-02-15T06:05Z 17.3K followers, 28.3K engagements

"10 distributed concepts we must know (one-liners): [--]. Timeouts: every network call needs a budget or it becomes a thread leak. [--]. Retries: they amplify load; use backoff + jitter and cap attempts. [--]. Idempotency: same request twice should not double-charge or double-write. [--]. Rate limiting: protect downstreams; fairness beats best effort under spikes. [--]. Circuit breakers: fail fast when a dependency is sick then probe recovery. [--]. Backpressure: when consumers are slow producers must slow down or drop. [--]. Consistency: pick the contract (strong vs eventual); document what users see. [--]. Leader"
X Link 2026-02-15T09:19Z 17.3K followers, 14.8K engagements

"I remember the feeling. 20182022: Coding was pure fun. CP late nights chasing that one AC. Every WA was a clue. Every AC was a small high. It felt like my brain vs the problem (and the world). Then you start working. [--]. The bugs arent in your code. Theyre in the requirements. [--]. The hard part isnt the algorithm. Its the migration with zero downtime. [--]. The win isnt AC. Its p99 going from 900ms to 120ms and nobody noticing. Still fun. Just different. If you miss the old feeling build a tiny thing again on weekends. No Jira. Just ship. Just ship for the love of it. 20182022: Coding was pure"
X Link 2026-02-15T09:21Z 17.3K followers, [----] engagements

"10 AI concepts to learn if youre trying to be a prompt engineer in 2026: [--]. Tokenization (why shorter isnt always cheaper) [--]. Context window + truncation [--]. System vs developer vs user messages (priority rules) [--]. Temperature top_p seed (controlling variance) [--]. Function calling structured outputs JSON schema [--]. RAG basics (chunking embeddings retrieval quality) [--]. Grounding + citations (dont trust fluent text) [--]. Evaluation (golden sets pass rate regression tests) [--]. Safety + prompt injection (your app will get attacked) [--]. Latency and cost (p95 retries rate limits) [--] distributed concepts"
X Link 2026-02-15T09:24Z 17.3K followers, [----] engagements

"10 DevOps concepts you must know for debugging (not for interviews): [--]. Structured logs + correlation IDs [--]. Metrics: RED/USE p95/p99 error budgets [--]. Distributed tracing (spans baggage trace sampling) [--]. Linux basics: CPU mem disk fd limits OOM killer [--]. Networking: DNS TLS NAT timeouts retries [--]. Containers: image vs runtime cgroups namespaces [--]. K8s: liveness/readiness rollouts events resource limits [--]. CI/CD: what changed what shipped how to rollback fast [--]. Incident hygiene: runbooks on-call handoffs postmortems [--]. Blast radius thinking: rate limits circuit breakers feature flags Always"
X Link 2026-02-15T09:25Z 17.3K followers, [----] engagements

"Your load balancer is a single point of failure. When it goes down your entire site is unreachable. How will you make load balancing highly available Asked at AWS interview https://twitter.com/i/web/status/2022974918586163499 https://twitter.com/i/web/status/2022974918586163499"
X Link 2026-02-15T10:03Z 17.3K followers, 113.4K engagements

"You have [--] microservices and clients need to call [--] different services to load a single page making [--] separate HTTP requests. How will you reduce client-side complexity Asked at Netflix interview https://twitter.com/i/web/status/2023035568548942100 https://twitter.com/i/web/status/2023035568548942100"
X Link 2026-02-15T14:04Z 17.3K followers, 64K engagements

"Ishan Kishan is on fire. Imagine you are working at jio hotstar. How would you ensure max uptime for the duration of this match where traffic only goes up What system components would you scale up/down for the next few hours"
X Link 2026-02-15T14:07Z 17.3K followers, 24.2K engagements

"As a future backend staff engineer. Please learn: - Advanced DB concepts with tradeoffs - Advanced System Design - Advanced API design - Advanced API Security - Advanced CI/CD - Basic Docker/ Kubernetes - Better communication with stake holders As a backend engineer. Please learn: - DB - System Design - Algorithms - API design - API Security - CI/CD - DS - Docker/ Kubernetes - Caching - Messaging Stop jumping from one language to the other As a backend engineer. Please learn: - DB - System Design - Algorithms - API design - API Security - CI/CD - DS - Docker/ Kubernetes - Caching - Messaging"
X Link 2026-02-15T14:10Z 17.3K followers, 23.8K engagements

"The [--] cities having most software devs shipping code fighting incidents and living in Git. [--]. Bangalore [--]. Hyderabad [--]. Pune [--]. Chennai [--]. Gurugram [--]. Mumbai [--]. Delhi [--]. Singapore [--]. London [--]. Berlin [--]. Amsterdam [--]. Dublin [--]. Stockholm [--]. Paris [--]. Toronto [--]. Vancouver [--]. New York City [--]. Seattle [--]. San Francisco Bay Area [--]. Austin 🍆 The [--] cities having the most sex [--]. Macau [--]. Krakow [--]. Guadalajara [--]. Sao Paulo [--]. Luxembourg [--]. Porto [--]. Marseille [--]. Bangkok [--]. Rio de Janeiro [--]. Panama City [--]. Cambridge [--]. Naples [--]. Medellin [--]. Brussels [--]. Hanoi [--]. 🍆 The [--] cities having"
X Link 2026-02-15T17:28Z 17.3K followers, 13.4K engagements

"You think you're struggling And you make excuses for not working hard. Be Jim Simons. Graduated MIT at age [--]. + PhD from Berkeley at age [--]. + Cracked Soviet codes for the NSA during the Cold War. + Fired for writing an anti-war letter to the New York Times. + Became Chair of the Math Department at Stony Brook at age [--]. Won the Veblen Prize in Geometry (the Nobel of Geometry). + Decided at age [--] that academic life was too slow. + Left his prestigious tenure at Stony Brook to strip-mall office in Long Island. Founded Renaissance Technologies with zero financial experience. Then the tragedy"
X Link 2026-02-15T17:48Z 17.3K followers, [----] engagements

"Having worked as a backend developer for almost [--] years I only recently realized the advantages of using cookies over local storage. If we store auth tokens in localStorage any XSS bug is simply game over. A single injected script can read our token and steal the account. Thats why FE folks prefer HttpOnly cookies for login sessions. - Browser sends it automatically - JS cant read it (harder to steal) - We can lock it down with Secure + SameSite Cookies do need CSRF protection but overall this is the safest default for web auth. Have you been noticing something for a long time but only"
X Link 2026-02-15T17:59Z 17.3K followers, 26.3K engagements

"As a Senior Backend Engineer how many concepts can you clearly explain (and actually implement) from below: [--]. Event-Driven Architecture [--]. Saga Pattern (choreography vs orchestration) [--]. CQRS (Command Query Responsibility Segregation) [--]. Event Sourcing (and when NOT to use it) [--]. Circuit Breaker Pattern (plus retries timeouts bulkheads) [--]. Distributed Tracing (trace IDs spans baggage sampling) [--]. CAP Theorem (and what it means in real systems) [--]. Idempotency (keys dedup exactly-once illusions) [--]. Data Sharding (routing rebalancing hot partitions) [--]. API Gateway (auth rate limits routing"
X Link 2026-02-16T05:43Z 17.3K followers, 102.2K engagements

"We keep seeing CDNs as the solution for caching static content. But what is a CDN and how do you implement one"
X Link 2026-02-16T07:28Z 17.3K followers, 17.6K engagements

"Your CDN caches user profile pages but users complain they see stale data after updating their profile. How will you invalidate cache efficiently Real issue at Twitter"
X Link 2026-02-16T10:50Z 17.3K followers, 11.3K engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing