@ConsciousRide Avatar @ConsciousRide Akshay Shinde

Akshay Shinde 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 15.32% stocks 10.81% finance 7.21% countries 3.6% social networks 3.6% celebrities 0.9% financial services 0.9%

Social topic influence redis #3, ai 9.91%, if you 9.01%, systems #90, realtime #919, design #790, in the 4.5%, data 4.5%, strong 4.5%, app 4.5%

Top accounts mentioned or mentioned by @0xlelouch_ @0xlelouch @sgm6912 @soroushp4 @dhanpalraj1998 @codelune259 @debsourya005 @ryancooper8311 @helderbuilds @shivemnn @asolovichh @saen_dev @varletint @sidjain_80 @eoeboh @sidjain80 @adityashinde @vivoplt @tawakkalah1310 @ashtwtz

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

Top Social Posts

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

"As an Engineer do you know if ChatGPT uses WebSockets/sockets for live chat streaming or just normal HTTP API calls with SSE #OpenAI #Engineering https://twitter.com/i/web/status/2019833244330684871 https://twitter.com/i/web/status/2019833244330684871"
X Link 2026-02-06T17:59Z [----] followers, 96.1K engagements

"50k payments per second with zero double charges is one of the hardest problems in backend engineering. One bug here means real money disappearing or charging customers twice. Idempotency first Client creates a UUID as idempotency key and puts it in the header of every payment request. Server checks key in redis or db table before processing. If it exists return stored response. If missing process payment once then store key and result. Retries just execute once. Write key first with unique constraint. Concurrent duplicates mean one processes and others return the existing outcome. Latency"
X Link 2026-02-10T12:08Z [----] followers, 18.5K engagements

"@CodeLune259 official flutter docs are the best starting point - for adv topics: net ninja flutter playlist on YT https://codelabs.developers.google.com/cat=Flutter https://docs.flutter.dev/ https://codelabs.developers.google.com/cat=Flutter https://docs.flutter.dev/"
X Link 2026-02-14T19:18Z [----] followers, [--] engagements

"As a frontend engineer please learn: - component composition over prop drilling - state management that doesn't turn into spaghetti - accessibility isn't optional (ARIA keyboard nav) - performance (lazy loading code splitting memoization) - security (XSS CSP sanitization) - async patterns and error boundaries - responsive design without media query hell - testing that actually catches bugs As a backend engineer. Please learn: - SOLID design principles - Multithreading - Immutability - Streaming messaging - Caching - Security - SSL JWT OAuth - factory decorator singleton obeservable design"
X Link 2026-01-25T04:36Z [----] followers, 111.8K engagements

"Frontend Roadmap Beyond React/Vue/Svelte Basics (2026 Edition) If you skip these you'll ship pretty UIs but not fast resilient accessible experiences that survive real traffic and angry users: [--]. Browser internals: event loop rendering pipeline critical rendering path layout/reflow/repaint triggers compositor vs main-thread work [--]. Performance reality: Core Web Vitals (p75/p95 LCP INP CLS) code splitting + prefetch lazy hydration virtualization (react-window / virtuoso) bundle analysis + treeshaking wins [--]. Accessibility (a11y) as engineering: semantic HTML + ARIA patterns focus traps &"
X Link 2026-01-27T01:42Z [----] followers, [----] engagements

"As an Engineer you should know the following concepts Idempotency Prevents duplicate payments Pagination Keeps your DB from dying Versioning Lets you evolve APIs without breaking clients Rate limiting Protects your service from abuse Error codes Helps clients handle failures correctly Caching Reduces load by 90% JWT security Prevents leaking sensitive data N+1 queries The difference between 10ms and 10s response time Docs Stops the how does this API work Slack messages Consistency Choose the right tradeoff for your use case"
X Link 2026-01-28T09:25Z [----] followers, 62.3K engagements

"Rate limiting and throttling both control how much someone can use your API but they handle it differently. Here's what actually separates them. Rate limiting blocks requests Rate limiting sets a hard cap on how many requests you can make in a time window. If you hit [---] requests per minute and try a 101st the server rejects it with a [---] status code. You're blocked until the window resets. The request doesn't happen at all. Throttling slows requests down Throttling doesn't reject requests outright. Instead it delays them or reduces their processing speed. If you exceed your limit the server"
X Link 2026-01-28T17:03Z [----] followers, 15.8K engagements

"A billion users sending messages every second means billions of read receipts firing constantly. One boolean column sounds simple until your database melts from write load. Here's how messaging platforms actually build this feature at scale == Why the naive approach fails Adding an is_read boolean to your messages table seems obvious. When someone reads a message flip it from false to true. But at scale this creates a write for every single message read. If a group chat has [---] members and one person sends a message that's [---] writes when everyone reads it. Multiply that across billions of"
X Link 2026-01-29T11:16Z [----] followers, 61.5K engagements

"If you cant explain these to a junior in plain words you probably dont fully get them yet. Connection pooling reuses existing database connections instead of creating new one every time. Saves time and stops the db from getting flooded with too many connections. Thread pools keep a set number of threads ready to handle tasks. Async processing lets you handle slow operations without blocking the thread so more requests get processed at once. Backpressure makes the system signal upstream to slow down when it's getting overloaded. Rate limiting caps how many requests are allowed in a given time"
X Link 2026-01-29T11:57Z [----] followers, [----] engagements

"Frontend Security [---] I wrote a complete frontend security hardening guide covering: [--]. Always use HTTPS (enforce HSTS + upgrade insecure requests) [--]. Set strong security headers (CSP X-Content-Type-Options X-Frame-Options Permissions-Policy Referrer-Policy) [--]. Sanitize & validate ALL user input before rendering (never trust client-side only) [--]. Prevent XSS properly (use DOMPurify / sanitize-html avoid dangerouslySetInnerHTML) [--]. Store tokens securely (HttpOnly + Secure + SameSite=Strict cookies over localStorage when possible) [--]. Implement CSRF protection (tokens for state-changing actions"
X Link 2026-01-29T15:48Z [----] followers, 27.9K engagements

"As a JavaScript dev how many of these [--] modern features can you comfortably use / explain [--]. Optional chaining (.) + Nullish coalescing () [--]. Logical assignment operators (&&= = =) [--]. Numeric separators (1_000_000) [--]. Top-level await in modules [--]. Private class fields (#private) & methods [--]. Promise.any() + AggregateError [--]. + [--]. Object.hasOwn() [--]. Temporal proposal stages (or at least basics if you've experimented) [--]. RegExp /v flag + RegExp.escape() (ES2025) Reply with your /10 http://Temporal.Now http://String.prototype.at http://Array.prototype.at http://Temporal.Now"
X Link 2026-01-29T18:31Z [----] followers, [----] engagements

"File uploads are one of the easiest ways attackers break into systems. Users can upload malicious code disguised as images executable files that exploit servers or bombs that consume all your storage. Here's how you actually secure file uploads in production. Never trust the file extension Validate MIME types from content Scan files for malware Limit file sizes Restrict file types explicitly Store files outside the web root Use random filenames Separate storage domain Set correct content type headers Disable script execution in upload directories Sanitize image uploads Rate limit uploads"
X Link 2026-01-30T01:03Z [----] followers, 23.4K engagements

"As a TypeScript dev how many of these [--] features do you actively use or understand well [--]. const assertions + as const [--]. Template literal types [--]. infer in conditional types [--]. keyof typeof + indexed access types deeply [--]. Utility types (OmitThisParameter ThisParameterType etc.) [--]. satisfies operator [--]. const type parameters (const T) [--]. Branded types / nominal typing patterns [--]. Decorators (stage [--] / experimental) [--]. Isolated declarations (TS 5.5+) / --isolatedDeclarations https://twitter.com/i/web/status/2017304555315167608 https://twitter.com/i/web/status/2017304555315167608"
X Link 2026-01-30T18:31Z [----] followers, [----] engagements

"Cloudflare dealing with 12h backups on 50TB ClickHouse here's how to cut that time hard without fancy tricks: Full backup every time is consuming the [--] hours. Switch to weekly full snapshot & daily incremental using clickhouse-backup. For append-only OLAP workload like this incremental only grabs new or changed partitions so backup time drops a lot often to couple hours or less. On restore side bring up recent data first; say last 7-30 days partitions to make the service live again quickly. Then let older data backfill async in background. Partition-level restore is what actually saves time"
X Link 2026-01-31T16:48Z [----] followers, [---] engagements

"Ever faced slow pagination after page [--] when dataset hits millions Offset & limit kills performance because database scans everything before the offset each time. Query gets slower as users scroll deeper. Switch to cursor-based instead. Pass last item's id or timestamp as cursor in next request. Backend does WHERE id last_cursor ORDER BY id ASC LIMIT [--]. Index on that column keeps it fast and consistent. No deep skips no duplicate rows if inserts happen mid-scroll scales way better. We moved our feeds and search results to cursor after offset started timing out at page 50+. Never looked"
X Link 2026-01-31T17:10Z [----] followers, [----] engagements

"common interview trap right now: fast in usa extremely slow in india; identical app backend devs usually miss the actual reason page is fast in usa slow in india same code same backend. distance is the real issue here. packets travel much farther to india so latency jumps. use CDN. put static files images css js closer to users. many places still think just optimize backend or compress more. no. network round trips decide this one. CDN fixes it very well. edge locations handle the geography part. thats the answer interviewers want. One common question that I am seeing in every interview"
X Link 2026-02-01T16:22Z [----] followers, [----] engagements

"8 things frontend devs should know in [----] (React + TS focused) (not the usual hooks tutorial stuff) [--]. Contracts first: Zod schema for props + api responses infer types Component can't be broken by bad data. [--]. Colocation over folder obsession: feature file/folder contains component + query + styles + tests Less jumping easier to reason about. [--]. No waterfalls: parallel preload or Suspense + async components Page feels slow mostly because of chained fetches. [--]. Accessibility basics matter: semantic tags keyboard nav quick screen reader pass Companies notice when it's ignored. [--]. Optimistic"
X Link 2026-02-01T17:30Z [----] followers, 18.7K engagements

"@0xlelouch_ nope in Go we dont hv anything that is 100% zod (fluent chaining + infer + parse/coerce all in one) but [--]. go-playground [--]. zog (OS - [--]. gozod (OS- these are pretty close http://github.com/kaptinlin/gozod http://github.com/Oudwins/zog http://github.com/kaptinlin/gozod http://github.com/Oudwins/zog"
X Link 2026-02-01T17:49Z [----] followers, [---] engagements

"We had users scrolling for [--] minutes and then waiting [--] seconds for the next page to load. Engagement dropped so low before we figured out what was actually breaking. Why offset breaks at scale OFFSET [-----] LIMIT [--] makes the database scan [-----] rows just to throw them away. The deeper you scroll the worse it gets. At [-----] items in queries take actual seconds. Use cursor pagination instead: Send the ID of the last item you showed. Next request asks for items after that ID. Database jumps straight there no scanning. Works the same on page [--] or page [---]. Pre-build feeds in background: Don't"
X Link 2026-02-01T18:23Z [----] followers, [----] engagements

"Active now on Instagram is not constant heartbeats or server-side timers ticking every second. It's smarter event-driven presence with a simple timestamp check. Client sends lightweight signals only when needed server pushes updates over WebSocket and green dot shows if last activity falls inside a short window like 30-60 seconds. App sends presence events mainly on key moments foreground open user taps screen sends message or scrolls feed. No spam every few seconds. Server records last_active timestamp for that user. Nothing fancy just one field updated on those events. When your app opens"
X Link 2026-02-02T12:13Z [----] followers, [----] engagements

"8 things fullstack devs should know in [----] (not another Next.js tutorial list) [--]. SSR vs CSR hydration tradeoffs: Controls first paint SEO interactivity timing. Know streaming and selective hydration rules. [--]. State that survives refresh and scale: Client state with Zustand or Jotai. Server state via tRPC or React Query. Avoid context soup. [--]. Database indexing when load hits: Compound indexes covering queries reading EXPLAIN. [---] ms to [--] ms matters at scale. [--]. API paradigm tradeoffs: REST versus GraphQL versus tRPC versus RSC. Understand over-fetching N+1 mobile battery impact. [--]. Auth"
X Link 2026-02-02T16:44Z [----] followers, [----] engagements

"Level it up further: Set up a simple CI/CD pipeline (GitHub Actions is free and painless) for auto-deploys on push Hook in structured logging + request tracing (OpenTelemetry basics or Sentry + something like Grafana/Prometheus if you are feeling ambitious) Add Redis or a CDN for caching auto-scaling rules if traffic spikes and at least one alert (p95 latency 200ms Slack ping) Bonus mastery: add basic cost monitoring (cloud billing alerts) and a health-check endpoint so the app can self-heal or roll back on failure https://twitter.com/i/web/status/2018365509855768849"
X Link 2026-02-02T16:46Z [----] followers, [---] engagements

"Full-stack Development Project Ideas Real-time Collaborative Editor E-commerce Platform with Stripe Social Media Dashboard Project Management Tool Video Conference Application Multi-tenant SaaS Platform Progressive Web App with Offline Mode Server-Side Rendered Blog with CMS Real-time Analytics Dashboard Food Delivery Platform Calendar and Scheduling System Learning Management System Invoice and Payment Tracking App Inventory Management System Customer Support Ticketing System Backend Development Project Ideas Authentication API CRUD API For Users Blog Backend URL Shortener Pagination and"
X Link 2026-02-02T17:05Z [----] followers, [----] engagements

"As a frontend developer Building real-time chat apps how do you handle the messages list Do you re-render the whole conversation component on every new msg Or use virtualized lists (react-window) memoized bubbles immutable state updates + auto-scroll logic Best practices for sender/receiver bubbles & performance #webdev #react #frontend https://twitter.com/i/web/status/2018468736018284864 https://twitter.com/i/web/status/2018468736018284864"
X Link 2026-02-02T23:37Z [----] followers, 17.6K engagements

"No primary key cannot be null. It has to be NOT NULL by rule in every proper database. Primary key = unique + not null. If null allowed then how will you identify the row properly NULL is unknown value so it breaks the whole point of having a primary key. One small exception people sometimes mention is composite primary key where one column can have null in very old systems or non-standard setups but in standard SQL and all major databases like PostgreSQL MySQL SQL Server Oracle; single column or composite primary key columns are forced not null. So answer is very clear that no it cannot be"
X Link 2026-02-03T04:33Z [----] followers, [---] engagements

"Compute the running balance when you write each transaction. Just add or subtract to previous balance and store it right there in the row. Index properly on user_id plus created_at or id. Reads become range scans instead of full table nonsense. Always paginate hard. Cursor based. Limit [--] or [---] rows max per call. Client asks for next page when needed. Never try to compute SUM over 1M rows on read. That kills postgres or any db under load. If you ever need to fix historical balances which is rare run a background job with window function on chunks not the whole table at once. Thats how ledger"
X Link 2026-02-04T18:43Z [----] followers, [----] engagements

"You're designing Twitter/X-style rate limiting for posting tweets / API calls (both per-user and global abuse prevention). How do you implement something that survives sudden viral spikes without over-limiting good users"
X Link 2026-02-05T12:11Z [----] followers, [----] engagements

"@0xlelouch_ Good call on Redis and token bucket. Redlock is popular but has some known gotchas around timing/clock sync that Kleppmann post from [----] explains it well. Still fine for most use cases though. https://martin.kleppmann.com/2016/02/08/how-to-do-distributed-locking.html https://martin.kleppmann.com/2016/02/08/how-to-do-distributed-locking.html"
X Link 2026-02-05T15:37Z [----] followers, [--] engagements

"@EOEboh Forward proxy is on the client side and hides IP from the sites that we visit. Reverse proxy is on server side and hides the backend servers from everyone connecting in. Direction is the difference. Cloudflare and Nginx handle the reverse one every day"
X Link 2026-02-06T17:16Z [----] followers, [--] engagements

"@SidJain_80 100% that's how they do it they dont use sockets and they stream via https + sse for simplicity streaming chunks and it is unidirectional do you know how Server Sent Events actually work under the hood"
X Link 2026-02-06T18:10Z [----] followers, [----] engagements

"Redis for increments no question. Just bump the count there for every view. Super fast no locking mess. Then a simple worker pulls the delta every few seconds or minute and adds it to the posts table in batches. One write instead of a million. CPU drops immediately. For the really viral ones you can shard the counter across a few keys so no single hot spot forms. Eventual consistency works perfectly here nobody cares if the number is exact to the second. Thats the practical fix Ive seen work at this scale. https://twitter.com/i/web/status/2020078632178467125"
X Link 2026-02-07T10:14Z [----] followers, [----] engagements

"@debsourya005 haha they are yet to make profit and if they go with sockets they can only dream of making profit 💀 SSE saving the day and the balance sheet"
X Link 2026-02-07T11:13Z [----] followers, [---] engagements

"We ran into this exact double charge issue once. Network retry fired the payment call twice. Same order got charged twice. Explaining refunds to pissed off customers is really irritating. Client side just generates a unique idempotency key uuid is fine and puts it in the header on every payment request. Server side first checks if that key already exists in Redis or a simple db table. If it exists return the stored response immediately. No reprocessing at all. If not found run the payment logic normally then save the key together with the full response or at least the success details. Set a"
X Link 2026-02-07T11:16Z [----] followers, [----] engagements

"@aditya__shinde SSE same as the regular completions/chat streaming apis. their official docs mention streaming task events over http + sse https://developers.openai.com/codex/ https://developers.openai.com/codex/"
X Link 2026-02-07T11:46Z [----] followers, [---] engagements

"Parking lot and movie ticket booking appear almost every time. Rate limiter forces you to think about token bucket vs fixed window vs sliding window and distributed setup with redis. Logging looks simple but people forget levels appenders async rotation correlation ids. Elevator needs good state handling for direction requests per floor priority. Vending machine and url shortener also come up often in follow-up questions. Practice explaining the class diagram key methods and one or two trade-offs clearly. https://twitter.com/i/web/status/2020114352913936777"
X Link 2026-02-07T12:36Z [----] followers, [---] engagements

"@vivoplt Positions that look safer right now are system thinkers who decide what should be built people who combine multiple agents and tools without constant checking and those who solve problems companies still pay serious money for even when code becomes very cheap"
X Link 2026-02-07T14:40Z [----] followers, [----] engagements

"A very popular post gets millions of likes/comments in minutes (hot key problem). Remember when Virat Kohli dropped that T20 World Cup victory post and it got like 20+ million likes in no time How do you stop your like-counter service (or database) from completely dying under that insane traffic spike What tricks do you use to handle these 'hot key' moments without everything crashing"
X Link 2026-02-07T18:28Z [----] followers, [----] engagements

"This happens because the API trusts the user_id from the URL instead of checking who actually owns the resource. The right way is to pull the user id straight from the auth token (jwt/session) and compare it with the resource owner on the server side. If they dont match just return [---]. Never trust anything coming from the client in these cases"
X Link 2026-02-08T08:10Z [----] followers, [----] engagements

"Full-Stack [--] Projects List Advanced Task Manager (Kanban + Drag & Drop + PWA) Personal Finance Tracker + Visual Dashboard Multi-Author Blog / CMS with Markdown Editor Real-time Collaborative Notes / Whiteboard URL Shortener with Analytics + QR Codes Recipe Finder + Meal Planner Event Management + RSVP + Maps Integration Job Board / Freelance Marketplace Full E-commerce Store (Cart + Payments + Admin) Mini Social Media (Posts + Likes + Threads + Feed) Learning Platform / Mini LMS (Courses + Quizzes) Appointment Booking System + Calendar Sync Habit Tracker with Social Streaks & Leaderboard"
X Link 2026-02-08T08:13Z [----] followers, 13K engagements

"@0xlelouch_ shipping something every month even tiny keeps you finishing instead of polishing forever writing weekly posts forces clearer thinking even when its simple start with one small thing today waiting for perfect moment usually turns into nothing at all"
X Link 2026-02-08T14:32Z [----] followers, [----] engagements

"Sliding window counters on Redis usually win here. Token bucket allows more bursts but sliding gives better fairness across the minute. For per-user and global you run two keys: one per user id and one shared global counter. Lua script does the increment and check atomically so no race between pods. If redis is single point then add replicas with eventual sync or go gossip style like envoy does but most places stick to redis and lua. For dynamic tiers just fetch limit from config service per request or cache it. Multi region add redis cluster per region and accept slight over limiting during"
X Link 2026-02-08T17:57Z [----] followers, [----] engagements

"This happens because name got cached somewhere even after DB update finished. Cache can sit in these places usually - browser local state or http cache headers - cdn edge layer - api gateway or proxy - backend redis/memcache - frontend query library like react query / apollo - sometimes read replica delay if read goes to secondary To make sure it shows correct name right away for the person who just changed it - after successful update call clear or update the exact cache entry - for that user force read from primary db or skip cache once - use optimistic update in ui set new name immediately"
X Link 2026-02-08T18:08Z [----] followers, [---] engagements

"Backend is dumping entire table or massive join without any limit. First thing I do is talk to backend guy saying that please fix the API add proper pagination or filters. No frontend magic fixes 100MB properly. If they cant change it right now: - Stream the response if possible (fetch with ReadableStream) - Parse only chunks I actually need - Use virtualization so UI doesnt try to render everything But honestly most of these cases fail on network and parse time before UI even starts. Actual fix is always smaller payload from server. That is it"
X Link 2026-02-08T18:15Z [----] followers, 10.2K engagements

"Graceful shutdown catches SIGTERM stops accepting new ws connections on old pods but keeps existing ones alive till they close naturally or timeout. Rolling update in k8s with readiness probes does the rest new pods come up healthy first traffic shifts slowly old ones drain before termination. State in redis so reconnect just works client retries handle the rare drops without user noticing anything. That's how you avoid the mass disconnect storm at 100k"
X Link 2026-02-09T05:11Z [----] followers, [----] engagements

"First check slowlog to see which command and value size is actually consuming CPU. Then pull that 1M qps read path out of Redis completely short ttl local cache in the app or shrink the payload if possible. If Redis has to stay in the loop just copy the hot key to a few instances and balance the gets across them. https://twitter.com/i/web/status/2020769965591445741 https://twitter.com/i/web/status/2020769965591445741"
X Link 2026-02-09T08:01Z [----] followers, [---] engagements

"This is the classic hot key situation in Redis. Single thread means one key getting hit like that pushes CPU to the limit. First thing I check is whether the data even needs to stay that hot in Redis every single time. Put a small local cache right in the app layer for that product key. Even a simple per-instance LRU or just a concurrent map with soft ttl cuts the traffic to Redis by a huge margin. If the value changes slowly then replicas help spread the reads. For proper fix though shard the key somehow. Append a bucket or user/region hash suffix so it spreads across instances instead of"
X Link 2026-02-09T09:52Z [----] followers, [---] engagements

"Docker stays useful even for small projects. Kubernetes only when traffic gets serious or you run many services. Add Redis somewhere for caching it changes speed a lot. Postman works but Bruno or Insomnia feel lighter these days. Jira can consume time if the team is small. good list we can use these tools based on project size. https://twitter.com/i/web/status/2020806207741845703 https://twitter.com/i/web/status/2020806207741845703"
X Link 2026-02-09T10:25Z [----] followers, [---] engagements

"@Tawakkalah13_10 nginx as load balancer is common but it can become single point of failure itself if not setup in active-passive or active-active mode with keepalived or similar"
X Link 2026-02-09T10:38Z [----] followers, [--] engagements

"@0xlelouch_ Netflix case showed this: app logs commit before the actual durable write happens. Transaction rolls back after. Or replication lag hits reads go to stale replica. Lesson stays simple. Never trust logs alone. Always check data at rest"
X Link 2026-02-10T01:20Z [----] followers, [---] engagements

"UUIDs as PK make joins slow because they are [--] bytes and random. Indexes get fragmented cache misses happen more comparisons take extra time. Integers stay compact and sequential so everything stays faster. Common fix is use UUID only for external / API exposure. Keep bigint surrogate key inside the service for joins. Or switch to UUIDv7 if you need ordered values without losing uniqueness. We did this in one project that bigint for internal relations UUID just for public refs. Queries became way quicker on large tables. https://twitter.com/i/web/status/2021032330853482857"
X Link 2026-02-10T01:23Z [----] followers, [---] engagements

"As an Engineer As you know ChatGPT uses the HTTP + SSE combo for user interactions / streaming replies. Do you know how to actually implement it And how it works under the hood"
X Link 2026-02-10T02:09Z [----] followers, [----] engagements

"@ash_twtz this list is gives good start as a beginner in js"
X Link 2026-02-10T06:26Z [----] followers, [---] engagements

"@code_sagar principles of building reliable systems havent changed"
X Link 2026-02-10T09:39Z [----] followers, [---] engagements

"grpc is heavily used but almost always for internal service to service communication in microservices architectures not usually for public/client facing APIs. big production orgs like netflix uber dropbox square spotify tictok robinhoood airnb slack cisco etsy and list goes on. https://twitter.com/i/web/status/2021184197407474155 https://twitter.com/i/web/status/2021184197407474155"
X Link 2026-02-10T11:27Z [----] followers, [---] engagements

"The only frontend concepts you need to know (2026 edition) Component composition Props drilling vs Context vs Composition Hooks rules useState useEffect useRef useTransition useOptimistic useActionState custom hooks State management Zustand / Jotai (simple) vs Redux Toolkit (complex) Data fetching TanStack Query / SWR Suspense + loading states Server Components React Server Components Server Actions Streaming UI Suspense boundaries incremental rendering progressive hydration Performance Memo useMemo/useCallback lazy + Suspense Accessibility Semantic HTML ARIA roles keyboard nav screen reader"
X Link 2026-02-10T12:28Z [----] followers, [----] engagements

"You're implementing real-time notifications (not just polling) for a chat and social app. WebSockets vs SSE vs push notifications vs long-polling what mix do you actually pick at different scales and how do you handle reconnections and duplicates gracefully https://twitter.com/i/web/status/2021230675500609659 https://twitter.com/i/web/status/2021230675500609659"
X Link 2026-02-10T14:32Z [----] followers, [---] engagements

"@RyanCooper8311 Great points. Hourly recon and timestamp validation on idempotency keys makes a lot of sense"
X Link 2026-02-10T18:00Z [----] followers, [--] engagements

"@Siddcodes1 Namecheap hands down better renewal prices free privacy forever and way less upsell nonsense than GoDaddy. Porkbun is also good if you want super transparent pricing"
X Link 2026-02-11T08:49Z [----] followers, [---] engagements

"To fix it put a CDN like Cloudflare or Fastly or AWS CloudFront with points in India Mumbai region. Cache responses at edge so Indian requests never reach origin most times. If API allows edge compute or regional replication do that too. That drops time close to what Australia sees without touching code. Simple geography fix not code fix. https://twitter.com/i/web/status/2021543800154562710 https://twitter.com/i/web/status/2021543800154562710"
X Link 2026-02-11T11:16Z [----] followers, [----] engagements

"@0xlelouch_ always better to look at actual responsibilities take home after tax and what cost of living"
X Link 2026-02-11T11:20Z [----] followers, [---] engagements

"People are pointing at the spacex merger as a possible trigger for some internal shakeup but the posts from Ba and Wu stay pretty calm talking about next chapters and staying friends with the team. No public drama or direct shots at leadership. Compute still being the big limiter makes the timing feel off especially when Elon says its handled. Talent walking away at this stage could slow things down even if the mission stays the same. Curious to see if more names surface soon or if they start backfilling fast. What do you think is really behind the wave"
X Link 2026-02-11T14:03Z [----] followers, [----] engagements

"Check locks first. [--] seconds for single row update almost always means blocked by another transaction. Run pg_locks or show engine innodb status (depending on db) find who is holding the lock on that row or table. Kill if needed then see why that other transaction is taking so long. If no lock then explain analyze the query see if index on where column is missing or not used. https://twitter.com/i/web/status/2021589477018448156 https://twitter.com/i/web/status/2021589477018448156"
X Link 2026-02-11T14:17Z [----] followers, [----] engagements

"these days for core AI engineering skills (LLM APIs RAG agents evals deployment patterns) the amount of high-quality free/very-low-cost material is massive: - Official docs + cookbooks from OpenAI/Anthropic/Cohere - YouTube (Andrej Karpathy lectures gkcs old free videos full paper walkthrus) - GitHub repos with production-grade examples -Cheap Udemy/Zero-to-Mastery style courses during sales the $1500 premium is mostly buying: - time compression (200 h claim) - live debugging & personalized feedback - cohort peer group & accountability If you are already disciplined can build side projects"
X Link 2026-02-11T14:41Z [----] followers, [----] engagements

"@rauchg Yeah Id say use AI aggressively or maybe smartly/wisely but never let it become a full-time brain outsourcer. Stay aware"
X Link 2026-02-11T16:41Z [----] followers, [---] engagements

"@brave Maybe a better UX experience and more UI customization. I use [--] browsers mostly [--] along with Brave"
X Link 2026-02-11T17:56Z [----] followers, [---] engagements

"Yeah exactly that is the mature architectural answer. FE should almost never be doing heavy aggregation / filtering / grouping over 10k100k+ raw rows in real-time. the great pattern now rn architecturally as you mentioned - - Backend / analytics DB (clickhouse pinot rockset starrocks duckdb+wasm in some newer cases or even TimescaleDB / QuestDB for time-series) pre-aggregates / rolls up the data - Returns compact materialized views / aggregated buckets (e.g. [--] min [--] min hourly depending on zoom level) - Frontend gets [--] kb json payloads instead of [--] to [--] MB for which can use TanStack Table +"
X Link 2026-02-12T01:21Z [----] followers, [---] engagements

"@naval American AI labs push open access and sharing tech widely. Yet the top open models right now are mostly from China: Kimi K2.5 DeepSeek V3.2 Qwen [--] GLM-4.7 sitting high on public leaderboards for reasoning coding math. The difference stands out"
X Link 2026-02-12T04:07Z [----] followers, [----] engagements

"Most replies here says FE and they are partially correct AI handles UI components layouts and styling fast because the patterns repeat a lot and training data for JS/HTML/CSS is huge. Backend still needs more careful logic around data flow auth scaling; AI messes up there more often so people review heavier. Albeit when it comes to architectural things flow of the application tradeoff decisions and complex logic again AI fails on FE too. So FE wins on volume of use right now but neither side lets you fully trust AI on the hard stuff yet. https://twitter.com/i/web/status/2021934524670574797"
X Link 2026-02-12T13:08Z [----] followers, [----] engagements

"@SoroushP4 Bloom filters are perfect for this tiny memory superfast checks and the tiny false positive risk is usually acceptable for block enforcement nice idea"
X Link 2026-02-12T19:07Z [----] followers, [--] engagements

"You're building a complex dashboard with 1000+ rows that update in real-time. What is your go-to combo right now - TanStack Table + Virtualizer - AG Grid (and pray the license fits) - Just react-window + custom everything - "I gave up and went full server components" Drop your stack & why #react #frontend"
X Link 2026-02-11T23:22Z [----] followers, [----] engagements

"Infosys fresher / early level still gets around 4-10 lakhs depending on band the [---] lpa looks like an entry package from couple years back but not far off for support kind of role. Amazon SDE [--] usually gets 23-30 lakhs total (closer to [--] like mentioned) with oncall and design work matching. Razorpay backend at [--] lpa fits mid level fintech pay there. Uber engineer [--] around [--] to 80+ lpa with rsus is normal for that level scaling apis and p99 stuff is standard. US remote startup at 85k usd converts roughly to same as uber or bit more after tax but full ownership & hiring involvement pushes"
X Link 2026-02-12T13:05Z [----] followers, 34.3K engagements

"As an AI/ML engineer. Please learn: - Math & theory foundations (linear algebra probability & statistics optimization information theory) - Deep learning architectures (Transformers diffusion models CNNs / ViTs RNNs/LSTMs/GRUs mixture-of-experts) - Training & scaling laws (pre-training fine-tuning PEFT / LoRA / QLoRA RLHF / DPO / PPO distributed training: DDP / FSDP / DeepSpeed / Megatron) - Data handling at scale (data pipelines cleaning & deduplication tokenization strategies synthetic data multimodal datasets) - Evaluation & benchmarking (automatic metrics: BLEU/ROUGE/BERTScore/MMLU human"
X Link 2026-02-14T13:24Z [----] followers, 129.4K engagements

"Cloud platforms and what they are good for: [--]. AWS: everything you can think of deep ecosystem. [--]. Google Cloud: data ML strong infra. [--]. Microsoft Azure: enterprise integration Microsoft stack. [--]. DigitalOcean: simple droplets developer-friendly. [--]. Linode: affordable VPS straightforward setup. [--]. Render: easy app hosting less DevOps overhead. [--]. Railway: quick backend deployment simple configs. [--]. deploy close to users edge-friendly. [--]. Heroku: easy PaaS less control but fast start. [--]. Oracle Cloud: cost-effective compute options. http://Fly.io http://Fly.io"
X Link 2026-02-14T19:45Z [----] followers, [----] engagements

"AI / ML Development Project Ideas - Simple Linear Regression House Price Predictor - MNIST Handwritten Digit Classifier - Sentiment Analysis on Movie Reviews or Tweets - Spam Ham Email or Message Classifier - Image Classifier Cats vs Dogs or Flowers - Customer Churn Prediction Telco or Bank dataset - Movie Product Recommendation System Collaborative Filtering - Time Series Forecasting Stock price Weather Sales - Face Detection plus Landmark Detection pipeline - Text Summarization Extractive or Abstractive with transformers - RAG-based QandA over your own PDFs or documents - Fine-tune small"
X Link 2026-02-15T07:01Z [----] followers, [----] engagements

"Frontend rendering & asset loading usually takes way more time than the API call. Backend [---] ms seems good but browser still waits on JS parse style calc layout paint. Users feel the full chain not only server response. Measure with devtools network and performance tab or lighthouse instead of trusting API logs alone. That shows the real story. https://twitter.com/i/web/status/2022941258444902888 https://twitter.com/i/web/status/2022941258444902888"
X Link 2026-02-15T07:49Z [----] followers, [---] engagements

"Single load balancer going down means nothing works anymore. Put them across different AZs so one AZ failure doesn't destroy everything. AWS ELB already spreads nodes automatically no single box controls it. Add Route53 health checks on top if you want extra failover layer. That's the basic fix they expect in AWS interviews. https://twitter.com/i/web/status/2022991110579409196 https://twitter.com/i/web/status/2022991110579409196"
X Link 2026-02-15T11:07Z [----] followers, [----] engagements

"ngl comforts addictive. the real glow-up starts when you say yes to uncomfortable things - eyes closed"
X Link 2025-10-20T07:01Z [----] followers, [---] engagements

"Your $3000 bed asking AWS for permission to change temperature is the perfect metaphor for how far we've strayed from building resilient products. I think AWS outages should happen more often So people realise the true value of decentralised internet and local offline modes. I think AWS outages should happen more often So people realise the true value of decentralised internet and local offline modes"
X Link 2025-10-21T10:07Z [----] followers, [----] engagements

"The next internet wont be websites: itll be agents talking to each other. HTTP was for humans. A2A (Agent-to-Agent) is for machines. Soon agents will: Negotiate API access Exchange structured data Authenticate + transact autonomously OpenAIs AgentKit Googles AP2 and the Model Context Protocol (MCP) are the early signs. Next app wont have an API route. Itll have an Agent Interface"
X Link 2025-11-15T17:18Z [----] followers, [---] engagements

"most chat uis start lagging badly once you cross [---] [----] messages. every new text forces the whole list to re-render and it feels choppy fast. this is how you actually build one that stays smooth even at 10k+ messages and rapid incoming msgs. - use virtualization right from the start. react-virtuoso is the cleanest choice for chats because it supports inverted scrolling natively and handles resize + new items without jitter. react-window works too if you flip the list yourself. - store messages in reversed order inside state. newest message lives at index [--]. this matches how most chat uis"
X Link 2026-02-04T17:45Z [----] followers, [---] engagements

"@DhanpalRaj1998 Its not really just those three things. Theres certificate validation against the chain key agreement (not just exchange) and the finished messages to confirm nobody tampered midway"
X Link 2026-02-06T15:41Z [----] followers, [--] engagements

"The bug is usually inconsistent routing to different backend instances that aren't in sync. One pod/server has fresh data or healthy db connection so returns [---]. Another has stale cache / delayed replication / missing record so gives [---]. Third one might be having connection pool exhaustion or disk full or deployment half-rolled so throws [---]. Retries hit random instances again so things get worse instead of better. You see this pattern a lot when autoscaling kicks in or during rolling updates or when read replicas lag. Tracing the request to see which pod/ip answered each time usually shows"
X Link 2026-02-07T14:28Z [----] followers, [----] engagements

"That viral spike overloads systems quick. Twenty million likes in a few minutes will break your basic counters if you havent set things up smart. Breakdown of what works at scale: - Redis grabs every like increment upfront. It uses atomic INCR operations to avoid any locking issues or race conditions during the flood. This keeps things moving without bottlenecks. - The app pulls the count from Redis for display. Users get the up-to-date number without waiting even under heavy load. - A background process batches the changes and writes them to a persistent store like Postgres or Cassandra"
X Link 2026-02-09T10:31Z [----] followers, [---] engagements

"Just queue the email task right after creating the user record. Push it to Redis list or SQS or any simple queue you already run. Return [---] to the frontend instantly with check your inbox message. Worker picks it up later and sends. User never waits those [--] seconds flow remains fast. That is the standard move for this. https://twitter.com/i/web/status/2021723058650964041 https://twitter.com/i/web/status/2021723058650964041"
X Link 2026-02-11T23:08Z [----] followers, [----] engagements

"We built something similar for one of the client and honestly tier upgrades were the thing that kept causing problems. Token bucket in Redis with Lua scripts This combo handled around 100k requests per second for us without issues. Lua scripts let you check tokens and decrement in one atomic operation so you don't get race conditions. Tier upgrades have to happen instantly Someone upgrades their plan and they expect higher limits right away. Making them wait [--] minutes for cache to expire felt terrible. We ended up invalidating cache immediately whenever subscriptions changed. Adaptive"
X Link 2026-02-12T02:27Z [----] followers, [---] engagements

"I will rank all the four options by strongest to weakest in order. [--]. SDE [--] @ Uber: strongest overall pick for most strong engineers [--]. SDE [--] @ Atlassian: very close second if you want significantly better work-life balance [--]. EM @ VMware: reasonable only if you are already excited about people management and don't mind a slower technical trajectory [--]. VP @ Goldman Sachs: usually the weakest real choice for a core software engineer unless you genuinely want to pivot into finance / trading systems long-term https://twitter.com/i/web/status/2021790082328670467"
X Link 2026-02-12T03:34Z [----] followers, 11.6K engagements

"Yeah this is the correct way. Upload file first give back [---] Accepted straight away. Then push the task to a queue or event. Worker picks it up does resize compress thumbnail updates status in redis or db. Frontend either polls the status endpoint every few seconds or gets push via websocket/socket. No blocking the http request for [--] seconds user sees upload done instantly progress shows later. Sync processing in api path kills response time and feels slow always. Async keeps it snappy. https://twitter.com/i/web/status/2021884046314467695 https://twitter.com/i/web/status/2021884046314467695"
X Link 2026-02-12T09:48Z [----] followers, [---] engagements

"@RahulBagal__ love this stack how is clickhouse query perf when you are doing aggregations on 1000+ updating rows in real time"
X Link 2026-02-12T18:10Z [----] followers, [--] engagements

"Redis per-user block sets are fine for the average case most block [---]. But scale breaks when a high-follower account gets blocked by millions in that case fan-out cost on read becomes insane. Xs algo applies blocks/mutes super early in the pipeline (before heavy ranking) accurately to avoid per-request lookups at that volume. See their [----] open source write up on visibility filtering. https://twitter.com/i/web/status/2022011234212782503 https://twitter.com/i/web/status/2022011234212782503"
X Link 2026-02-12T18:13Z [----] followers, [--] engagements

"We ran into this puzzle at one of the product based companies and the honest answer is the mix really depends on the feature and scale. WebSockets for active chat If someone has the chat open and typing WebSockets make sense. Bidirectional so you can send typing indicators both ways. Connection stays alive as long as they're in the app. SSE for feed updates For things like notifications bell or activity feed SSE works fine. Server pushes updates to the client over regular HTTP. Way simpler than WebSockets when you only need one direction. Push notifications when app is closed If the app isn't"
X Link 2026-02-12T18:33Z [----] followers, [---] engagements

"Uber loves throwing variants like minimum/maximum value that satisfies X condition where you binary search over the possible range instead of the array itself. Examples: aggessive cows book allocation ship packages within D days split array largest sum capacity to ship packages. https://twitter.com/i/web/status/2022207688566886838 https://twitter.com/i/web/status/2022207688566886838"
X Link 2026-02-13T07:14Z [----] followers, [----] engagements

"@0xlelouch_ ai is convenient story to tell instead of admitting growth slowing and costs biting focus stays same that build real skills sharpen fundamentals ans ignore the loud narratives"
X Link 2026-02-13T10:19Z [----] followers, [--] engagements

"@0xlelouch_ Pick something real from Uber or Netflix pull out the key decisions and where they admitted things went wrong then build a small version yourself even if its just stubs and fake traffic. That one implementation teaches way more than saving twenty links you never revisit"
X Link 2026-02-13T12:33Z [----] followers, [---] engagements

"This approach works really well at scale. Storing blocks in something durable like Cassandra or DynamoDB & Redis cache for fast checks is standard. Kafka for async propagation to other systems prevents blocking the write path. For new posts skipping fanout to blocked/blocking users saves a ton of useless work on storage and compute. For older posts the runtime check during render/filter is the only accurate way since you can't rewrite history feeds retroactively. Many big platforms do exactly this combo. https://twitter.com/i/web/status/2022713080569040916"
X Link 2026-02-14T16:42Z [----] followers, [--] engagements

"For devs the message is clear. Pick one area that actually moves the business number payments infra security pipelines whatever it is and get really sharp there. General skill still matters but owning one metric people can point to cuts through faster in interviews. Use AI to move quicker on the boring parts dont make it the whole identity. Debugging system design and handling live incidents remain human edges for now. Build some real runway if you can. Six to twelve months breathing room changes how you think about offers and moves. Markets tighten selectivity goes up fundamentals win again."
X Link 2026-02-13T14:26Z [----] followers, [---] engagements

"@helderbuilds digital ocean clear win coz its way simpler & cheaper for early stage. aws will burn time and money until you have scale. switch to aws once you actually need the extra services"
X Link 2026-02-15T05:13Z [----] followers, [--] engagements

"The only AI/ML concepts you need to know (practical 2026) LLM basics Tokens Context window Temperature / Top-p Prompt engineering Chain-of-thought Few-shot Role prompting RAG Embeddings Vector DBs (Pinecone / PGVector) Retrieval + generation Fine-tuning LoRA / QLoRA PEFT methods Agents Tools ReAct loop Memory Workflows vs Chains Evaluation BLEU/ROUGE vs LLM-as-judge Human eval basics Inference optimization Quantization (4-bit/8-bit) vLLM / TensorRT-LLM Hallucination fixes Grounding Self-consistency Retrieval checks Deployment API endpoints Auto-scaling Monitoring (drift latency) SLM vs LLM"
X Link 2026-02-08T19:00Z [----] followers, 81.9K engagements

"Interviewed an Al engineer candidate last week with [--] years in ML for my friend's startup. Couldn't explain transformer architecture properly. Didn't know about RAG pipelines. Never worked with vector databases. But gave him a prompt engineering problem - his systematic approach to context optimisation and error handling was on point. Hired him. You can learn tools and frameworks. You can't easily teach logical thinking and problem-solving methodology. https://twitter.com/i/web/status/2020578242185638087 https://twitter.com/i/web/status/2020578242185638087"
X Link 2026-02-08T19:19Z [----] followers, 147.3K engagements

"The only backend concepts you need to know API paradigms REST GraphQL tRPC/gRPC trade-offs Authentication JWT vs Sessions OAuth2 flows Refresh tokens Authorization RBAC vs ABAC Middleware guards Databases SQL (Postgres default) vs NoSQL (when to choose) Caching Redis patterns (TTL cache-aside write-through) Rate limiting Token bucket Leaky bucket Sliding window Async processing Queues (RabbitMQ / Kafka) Background jobs Error handling Global exceptions Custom error classes Logging & Monitoring Structured logs Metrics (Prometheus) Tracing Scaling basics Horizontal vs Vertical Load balancers"
X Link 2026-02-09T15:59Z [----] followers, 30.7K engagements

"Design user blocking at scale (Twitter/Instagram level). A user can be blocked by millions of people; blocked users shouldn't see their content or interactions. How do you enforce this efficiently across feeds search recommendations"
X Link 2026-02-12T11:09Z [----] followers, [----] engagements

"Imagine building the future at Google and then waking up to an email that basically says: [----] was great. Now its AI or the exit door." This isnt a thought experiment. Its happening. Google is offering voluntary exit packages to parts of its U.S. business teams. Not layoffs. Not officially. Theyre calling it a 'voluntary exit program.' Translation: 'You can leave. Well pay you.' Last year similar packages reportedly included: [--] weeks base pay + [--] extra week for every year of service So if you have been there [--] years Thats approx [--] weeks of pay. Not terrible. But heres whats interesting"
X Link 2026-02-13T09:46Z [----] followers, [---] engagements

"You run a high-write counter system (views likes upvotes) that increments millions of times per minute. One day you realize a 64-bit unsigned int will overflow in 500+ years. Do you care What would make you add overflow protection anyway"
X Link 2026-02-13T10:25Z [----] followers, [----] engagements

"Kafka catches every comment and emoji at the edge split by match and user country so load spreads out. Flink runs fast spam checks and rate limits before stuff even hits the main database. For sending feeds out: Redis pub sub per region plus write time fan out to each user sorted set. That way every viewer pulls their stream super quick. Sentiment part runs a small distilled model something like MobileBERT size right on edge nodes near the big broadcast zones. Gets inference done really fast then pushes the score back through the same Redis channel. Main tradeoff: edge speed is great but the"
X Link 2026-02-14T13:30Z [----] followers, [---] engagements

"@shivemnn yes without strong math you are just memorizing architectures not understanding why they work. id say if you are on more of ML side then maths is must if you are just building agentic workflows llm integrations and all those stuffs then you can skip maths"
X Link 2026-02-15T04:43Z [----] followers, [---] engagements

"we need to get smarter about this and bridge the gap instead of fearing it. soft skills system/architecture thinking infra/ cloud depth scaling agentic aistandardising ai protocols; these are next ai proof things until ai reaches that level the move is always pick the next thing faster than it picks is. imo as longs as we keep evolving nothing can potentially replace us. https://twitter.com/i/web/status/2022905646656716860 https://twitter.com/i/web/status/2022905646656716860"
X Link 2026-02-15T05:27Z [----] followers, [--] engagements

"This makes complete sense right now. Traditional full stack or app dev paths are packed already freshers struggle to break in without experience. But agent side still has room because not every senior engineer jumps on it fast. Starting with small POCs in RAG or multi-agent setups gives actual proof you can build something useful. Startups grab people who show they can ship agent stuff early. Later big tech will chase the same talent when they shift hiring by then you already have real work to show. This is the direction for anyone in college today."
X Link 2026-02-15T05:34Z [----] followers, [---] engagements

"Popular JavaScript Frameworks / Frontend Libraries (20252026 trends) - React: United States - Vue: China (created by Evan You) - Angular: United States - Svelte: United States (New Zealand creator) - Next.js: United States - Nuxt: France - Astro: United States / Canada - Solid: United States AI / LLM-origin heavy hitters (publicly known base models 2025) - GPT series: United States - Llama series: United States - Claude: United States - Gemini: United States - Mistral: France - Grok: United States - DeepSeek: China - Qwen: AI / LLM-origin heavy hitters (publicly known base models 2025) - GPT"
X Link 2026-02-15T15:01Z [----] followers, [---] engagements

"As a frontend engineer Please master these fundamentals instead of chasing every new shiny library or framework every few months. Focus on going deep in - JavaScript and TypeScript core mechanics (event loop closures prototypes this keyword async patterns microtask queue) - Modern CSS (Flexbox and Grid mastery container queries cascade layers logical properties :has selector scroll-driven animations) - Responsive and adaptive design (mobile-first approach fluid typography with clamp smart breakpoints without media-query overload) - Browser rendering and performance (Critical Rendering Path"
X Link 2026-02-15T15:34Z [----] followers, 11.4K engagements

"Cart data remains in server memory. Next request goes to another server cart gone. Real fix is simple. Make servers stateless. Put cart in Redis (or DB if needed). Cookie holds only session ID. Every server pulls the same cart using that ID. Sticky sessions look easy but create single points of failure and uneven load. Avoid them in production. This is how scalable e-commerce actually runs. https://twitter.com/i/web/status/2022690562449764743 https://twitter.com/i/web/status/2022690562449764743"
X Link 2026-02-14T15:13Z [----] followers, [---] engagements

"No I dont care about overflow that [---] years away. At millions of increments per minute a uint64 still takes insane time to wrap. Way beyond any system lifetime we can imagine right now. I'd rather spend energy on code that is easy to read change and debug years later than add checks nobody will ever hit. Only thing that would push me to add protection is if the counter actually meant something critical past that point like money totals or irreversible IDs where wrap breaks logic hard. For plain metrics views/likes/upvotes Nope ignore it. https://twitter.com/i/web/status/2022716419079860371"
X Link 2026-02-14T16:55Z [----] followers, [--] engagements

"@ASolovichh thanks for the nudge lets keep it strict containers & DBs next time"
X Link 2026-02-15T19:19Z [----] followers, [--] engagements

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