@udaysy Avatar @udaysy Uday Yatnalli

Uday Yatnalli posts on X about claude code, ai, vibe coding, anthropic 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 31.29% finance 4.68% social networks 4.32% stocks 3.24% countries 2.52% vc firms 0.36% products 0.36%

Social topic influence claude code #343, ai 12.23%, vibe coding #453, anthropic #886, code #52, devs #398, agentic #1046, open ai 2.88%, products #3965, loops 2.88%

Top accounts mentioned or mentioned by @antirez @heygeorgekal @claudeai @robinebers @theo @anthropicai @yuchenjuw @melvynxdev @emollick @expenselens @fortelabs @camsoft2000 @danielmac8 @aaronontheweb @tannerlinsley @joeabbud @dkubb @cursorai @nummanali @simonw

Top assets mentioned Alphabet Inc Class A (GOOGL) Microsoft Corp. (MSFT)

Top Social Posts

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

"higher fps means the gpu is doing MORE work per second not less. the old path was probably bottlenecked on cpu or had idle frames where the gpu could cool down. now youre saturating the gpu pipeline continuously which = heat. might want a frame rate cap at like 60fps so you get the smoothness without cooking the machine https://twitter.com/i/web/status/2016173761104642291 https://twitter.com/i/web/status/2016173761104642291"
X Link 2026-01-27T15:37Z [---] followers, [---] engagements

"@vig_xyz the urgency gap is the real consulting moat rn"
X Link 2026-02-02T04:17Z [---] followers, [----] engagements

"@heygeorgekal hetzner + uptime kuma. took [--] mins to deploy monitors everything sends alerts to telegram. datadog pricing is designed for companies with vc money not indie hackers"
X Link 2026-02-03T19:08Z [---] followers, [--] engagements

"@julesagent supabase mcp changed how i work with claude code. db queries straight from the terminal without context switching. how does jules handle the mcp auth layer tho thats usually where integrations fall apart"
X Link 2026-02-03T22:06Z [---] followers, [---] engagements

"@Daviowhite run an ai-first dev agency. for design: v0 for rapid prototyping (skip figma-to-code entirely) midjourney for brand assets claude code for component iteration. went from making [--] layout to generating [--] variants in the same time. clients cant tell the difference"
X Link 2026-02-03T22:07Z [---] followers, [---] engagements

"@dylanscandinaro anthropic safety person heading to openai right when they are scaling to the next gen of models. timing says a lot about where the real deployment risk sits right now"
X Link 2026-02-03T22:11Z [---] followers, [---] engagements

"@ZackKorman npm supply chain attacks but for ai agents. the skills discovery being opt-out instead of opt-in is the core issue. sandboxing should be default"
X Link 2026-02-03T23:10Z [---] followers, [---] engagements

"@weswinder oh the vertex leak yeah saw some people mentioning that. if its real thats a fast turnaround from opus 4.5"
X Link 2026-02-04T00:19Z [---] followers, [--] engagements

"@rileybrown @AnthropicAI @vibecodeapp opus [---] definitely raised the bar but tbh the gap between "ships in demo" and "ships in production" is still huge. most people vibe coding arent handling edge cases auth flows or database migrations. the ones who ship real stuff already knew how to build before"
X Link 2026-02-04T00:33Z [---] followers, [---] engagements

"@camsoft2000 apple shipping an MCP server is the real signal here. when platform owners start adopting the protocol instead of building proprietary alternatives thats when you know the standard won"
X Link 2026-02-04T00:38Z [---] followers, [--] engagements

"@tobi been rolling my own memory setup with markdown files + session logs. having it native changes the workflow completely no more manual context stitching between sessions"
X Link 2026-02-04T03:30Z [---] followers, [----] engagements

"the source of truth distinction is the right framing. datadog and salesforce arent getting vibecoded away because nobody wants to own that infra and data pipeline themselves. but the lighter saas stuff weve already started replacing some of those internally. took maybe [--] weekends with claude code to build what we were paying $200/mo for. the real question is where exactly that line sits between "build it" and "just subscribe" https://twitter.com/i/web/status/2018893170139078921 https://twitter.com/i/web/status/2018893170139078921"
X Link 2026-02-04T03:43Z [---] followers, [---] engagements

"@_chenglou wait this is sick. been using ripgrep for everything but mdfind would catch stuff outside the project dir too right like config files scattered across /"
X Link 2026-02-04T05:57Z [---] followers, [---] engagements

"@0xSero @NoahGreenSnow points [--] and [--] are legit pain points. ran into the 2.5GB memory spike on a 4hr session last week. on point [--] tho codex for debugging and architecture been the opposite for me. opus catches design flaws in plan mode that codex misses entirely. maybe task dependent"
X Link 2026-02-04T08:22Z [---] followers, [--] engagements

"@davis7 the "just stick in the repo url" part is what sells it. been doing this manually with context7 mcp but having cli + web chat + mcp all in one package is way cleaner. how are you handling large monorepos tho thats where context tools usually choke"
X Link 2026-02-04T08:32Z [---] followers, [--] engagements

"@dr_cintas use elevenlabs for every video voiceover. v3 model with creative mode is genuinely hard to tell from a real person. $11B makes sense when the product actually works"
X Link 2026-02-04T17:29Z [---] followers, [--] engagements

"@Yuchenj_UW the second order effect is what gets me. once openai trains users to expect free + ads every paid tier becomes harder to justify. anthropic staying B2B and developer focused means they never have to optimize for attention. same playbook google should have kept with search"
X Link 2026-02-04T18:37Z [---] followers, [----] engagements

"@Miles_Brundage both companies spending energy on marketing drama instead of shipping feels like the exact mistake they tell startups not to make. meanwhile their actual users just want faster inference and better context handling"
X Link 2026-02-04T18:39Z [---] followers, [---] engagements

"@0xdoug the small teams who use these tools to ship products that would have needed [--] people. value accrues to the orchestrators not the infrastructure"
X Link 2026-02-04T18:40Z [---] followers, 13K engagements

"@mattshumer_ biggest thing ive learned running agent loops is the handoff protocol matters way more than the individual agent quality. does he cover how they handle state between agents"
X Link 2026-02-04T20:02Z [---] followers, [---] engagements

"codex nails the UI. but the second you need mcp servers custom hooks or skills that chain together youre back in the terminal. different tools for different builders The Codex macOS app is so much better than the Claude Code CLI. Might switch just for that. Now I just need a companion iOS app and Ill be a happy vibe coder. The Codex macOS app is so much better than the Claude Code CLI. Might switch just for that. Now I just need a companion iOS app and Ill be a happy vibe coder"
X Link 2026-02-04T20:18Z [---] followers, [---] engagements

"@estemdetornada @0xdoug distribution is the moat now not the build. everyone can build the same thing. the ones who win are the ones who get to market first and compound relationships while competitors are still scoping"
X Link 2026-02-04T21:22Z [---] followers, [---] engagements

"@TimJayas b2b every time as a solo founder. ran b2c for years and the support volume alone will eat you alive. b2b customers pay more churn less and you can close deals with a good demo instead of burning cash on ads"
X Link 2026-02-04T21:32Z [---] followers, [--] engagements

"@swyx @supabase the supabase comparison is spot on. same playbook too: absurdly good docs ship weekly make devs feel heard. anthropic just doing it with models + tooling instead of postgres"
X Link 2026-02-04T21:46Z [---] followers, [---] engagements

"@daniel_mac8 been switching between both. codex is faster for single file edits claude code wins on multi-file ops where it needs to hold context across 10+ files"
X Link 2026-02-04T23:16Z [---] followers, [--] engagements

"@arvidkahl been noticing this firsthand. the devs on my team who write better specs get dramatically better output from claude code than the ones who just say "build me a dashboard". same tool same model 3x difference in results. writing was always the bottleneck code just hid it"
X Link 2026-02-05T02:05Z [---] followers, [----] engagements

"@melvynxdev codex has a nice UI but the extensibility gap is massive. no MCP no hooks no skills. if you just want a chat that writes code its fine. if you want a programmable coding agent claude code isnt close to being matched yet"
X Link 2026-02-05T02:06Z [---] followers, [---] engagements

"every h1b restriction is a gift to bangalore google: 2.4M sqft new office meta: expanding india eng microsoft: largest office outside redmond us thought it was punishing indian workers it was actually funding their competitors back home built in india distributed globally. the playbook is writing itself. https://twitter.com/i/web/status/2019244589933092981 https://twitter.com/i/web/status/2019244589933092981"
X Link 2026-02-05T03:00Z [---] followers, [---] engagements

"@cryptopunk7213 paying customers ad revenue. the moment you optimize for engagement instead of output quality the product rots from the inside. anthropic gets this"
X Link 2026-02-05T04:27Z [---] followers, [---] engagements

"@alz_zyd_ same thing happening in software agencies. the junior dev role is getting compressed into AI tooling so every human on the project is expected to contribute senior level thinking"
X Link 2026-02-05T04:30Z [---] followers, [---] engagements

"the persistent local context is what keeps me on it. knows my projects patterns goals without re-explaining every session. setup was rough tho spent a full afternoon on gateway config. biggest wish: better model switching UX. editing config files every time i wanna swap providers feels like [----] https://twitter.com/i/web/status/2019301046766674319 https://twitter.com/i/web/status/2019301046766674319"
X Link 2026-02-05T06:44Z [---] followers, [---] engagements

"@Codie_Sanchez the 2am doom scroll through AI announcements is a lifestyle at this point. my approach: pick the 2-3 that solve problems you actually have rn and ignore the rest. the other [----] will either die or still be there next month"
X Link 2026-02-05T06:46Z [---] followers, [---] engagements

"@zeeg $96 for a single session is commitment lol. i feel the diff review gap end up doing git diff in a split pane which works but isnt great. codex nailing that UX while claude code has the deeper agentic loop feels like itll force anthropics hand on the visual side"
X Link 2026-02-05T09:34Z [---] followers, [--] engagements

"@nico_jeannen 2500+ ads from openai vs [----] from anthropic but their strategies are completely different. openai is fighting for consumer mindshare anthropic is targeting developers and enterprise. wonder which ad spend has better CAC"
X Link 2026-02-05T12:28Z [---] followers, [---] engagements

"dear @AnthropicAI "smooshing" is fun but when i have [--] terminal windows running claude code i genuinely forget what each one is doing feature request: show a one-liner of what the model is actually working on "working on auth flow" "smooshing" "refactoring database queries" "cogitating" would really help people with adhd like me who open terminals and immediately forget why https://twitter.com/i/web/status/2019397327707144323 https://twitter.com/i/web/status/2019397327707144323"
X Link 2026-02-05T13:06Z [---] followers, [---] engagements

"@Aaronontheweb @lxztlr @alxfazio workspaces would be huge. right now i manually partition memory across project folders and its messy. if v2 can handle cross project references too thatd basically replace my whole setup"
X Link 2026-02-05T15:04Z [---] followers, [--] engagements

"seen this firsthand working with indian IT companies. the whole business model optimizes for billing hours not building products. when your revenue scales linearly with headcount theres zero incentive to invest in AI that reduces headcount. the ones that might pull it off are the smaller companies that never went public"
X Link 2026-02-05T15:07Z [---] followers, [---] engagements

"@Param_eth im in this post and i dont like it. added docker to that hate list after the 47th time debugging container networking at 2am"
X Link 2026-02-05T15:16Z [---] followers, [----] engagements

"@trikcode theyre not the same skill. vibe coding is product thinking with AI as the execution layer. programming is understanding what the AI is doing so you can fix it when it breaks. you need both but the ratio is shifting fast"
X Link 2026-02-05T15:18Z [---] followers, [---] engagements

"honestly the tool doesnt matter as much as people think. chatgpt cursor claude code all get you 80% of the way. the gap isnt awareness its knowing what to build and when the output is wrong. ive seen devs with chatgpt ship faster than devs with cursor who dont understand the code https://twitter.com/i/web/status/2019455997849530833 https://twitter.com/i/web/status/2019455997849530833"
X Link 2026-02-05T17:00Z [---] followers, [---] engagements

"@s_ketharaman @deepakshenoy the accenture/capgemini parallel is spot on. services DNA and product DNA are fundamentally different organisms"
X Link 2026-02-05T18:05Z [---] followers, [--] engagements

"@weswinder the 0.1% swe-bench dip is noise tbh. look at the agentic computer use jump 72.7% vs 66.3%. thats where this actually matters. terminal coding benchmarks dont capture how much better the tool use + context window combo feels in actual workflows"
X Link 2026-02-05T18:16Z [---] followers, [---] engagements

"@github @AnthropicAI @code been running opus in claude code all day. the planning on multi-file refactors is noticeably better than [---]. havent tried it in copilot yet tho wonder if the agentic mode feels different there vs terminal"
X Link 2026-02-05T18:20Z [---] followers, [----] engagements

"been using gpt for backend and opus for frontend for months. the convergence you describe is exactly why the harness layer matters more now. if both models get roughly equal at everything the differentiator becomes the workflow built around them not the model itself"
X Link 2026-02-05T19:47Z [---] followers, [--] engagements

"@1Umairshaikh worst part is the "validation" is usually just talking to other founders who tell you what you want to hear. shipped [--] products first real signal came from someone pulling out a credit card not from a survey"
X Link 2026-02-05T19:53Z [---] followers, [--] engagements

"@RickBakas @supabase @claudeai right felt like deja vu reading it. the mcp server has been working great for a while now maybe they just made it official in the claude web/desktop UI"
X Link 2026-02-05T22:19Z [---] followers, [--] engagements

"@jessfraz the background terminals are genuinely nice. claude code has had something similar with the Task tool spawning subagents but codex making it a first class UX thing is smart. my main question is whether live steering actually helps or just creates more context pollution mid-task"
X Link 2026-02-05T22:23Z [---] followers, [---] engagements

"@badlogicgames the prompt adherence improvement is lowkey the most important part of this release. been using claudes equivalent (the project config file) for months and once the model stops ignoring your instructions the whole workflow changes. you go from babysitting to composing"
X Link 2026-02-05T22:36Z [---] followers, [---] engagements

"@shafu0x tmux + claude code agent teams is genuinely the move. four panes four agents zero context switching"
X Link 2026-02-05T23:50Z [---] followers, [--] engagements

"@code finally. managing [--] claude code agents across tmux panes works but this is way cleaner"
X Link 2026-02-05T23:59Z [---] followers, [---] engagements

"@melvynxdev 93% retrieval at 256k dropping to 76% at 1M is actually way better than i expected. most models fall off a cliff past 200k. the real question is whether it holds up when you're not just retrieving but reasoning across that full window"
X Link 2026-02-06T01:39Z [---] followers, [---] engagements

"@RhysSullivan the async model is what makes it click tbh. fire off [--] codex tasks keep vibing in claude code come back to finished PRs"
X Link 2026-02-06T01:40Z [---] followers, [---] engagements

"similar setup here. opus for anything that needs to understand the full UI tree and make design judgment calls codex for background tasks where you just want it done. the real win is running them simultaneously tho. codex churning through migration scripts while you're iterating on the UI with opus in real time https://twitter.com/i/web/status/2019588054168793140 https://twitter.com/i/web/status/2019588054168793140"
X Link 2026-02-06T01:44Z [---] followers, [---] engagements

"@ChristianSelig @SnazzyLabs same experience. codex wins on raw output speed but claude code's tooling is where it pulls ahead. the autonomous multi-file refactors git commits debugging loops without babysitting. the model is one variable the harness around it is what makes or breaks the workflow"
X Link 2026-02-06T03:31Z [---] followers, [---] engagements

"@cblatts seeing the same pattern. we use claude code for data pipeline stuff and the research assistants who adopted it fastest were the ones who already had domain knowledge. the tool amplifies what you already know doesnt replace the knowing"
X Link 2026-02-06T03:40Z [---] followers, [----] engagements

"@JanarJurisson both sides simultaneously. shipping a feature at 2am feels like the sunny side. waking up to a support ticket about it at 6am is the other side"
X Link 2026-02-06T03:47Z [---] followers, [--] engagements

"@Hesamation the notes in a folder pattern is basically what the claude md config does in claude code. you give the agent persistent context about the project so it doesnt rediscover things every session. retrieval is where most people skip and then wonder why their agent keeps forgetting"
X Link 2026-02-06T03:57Z [---] followers, [---] engagements

"@Aaronontheweb @lxztlr @alxfazio lol ship it. ill be the first guinea pig"
X Link 2026-02-06T04:56Z [---] followers, [--] engagements

"@ggganeshh @viprayami1 fair point. enterprise scale adds a whole layer of compliance integrations edge cases that vibe coding cant touch yet"
X Link 2026-02-06T07:41Z [---] followers, [--] engagements

"@TheAhmadOsman same. claude code turned my terminal into the main workspace. ide is basically a viewer now"
X Link 2026-02-06T07:46Z [---] followers, [---] engagements

"this is the use case people sleep on. everyone talks about ai writing code but using it as a second pair of eyes on existing codebases is where it genuinely saves weeks. i ran something similar on a 30k line next.js app and it caught [--] race conditions in supabase realtime subscriptions that wouldve taken forever to reproduce in prod. the trick is keeping the scope narrow like you did just bugs not rewrites"
X Link 2026-02-06T13:27Z [---] followers, [----] engagements

"been saying this for months. the harness matters more than the model. opus [---] was already capable but it felt like driving a sports car in first gear. [---] with the extended thinking budget just lets the same intelligence actually breathe. you can literally feel the difference in how it approaches multi file refactors https://twitter.com/i/web/status/2019788697080901915 https://twitter.com/i/web/status/2019788697080901915"
X Link 2026-02-06T15:02Z [---] followers, [--] engagements

"using it in claude code not cursor and the autonomous execution is where it shines for me. codex writes tighter functions but opus handles the "go refactor this entire module and update all the tests" tasks without losing context halfway through. different tools for different scopes https://twitter.com/i/web/status/2019791414264263158 https://twitter.com/i/web/status/2019791414264263158"
X Link 2026-02-06T15:12Z [---] followers, [---] engagements

"@theandreboso yeah fair. slack and notion pulled it off but they had years of runway to wait for conversion. most early stage founders dont"
X Link 2026-02-06T16:29Z [---] followers, [--] engagements

"@paul_popov openrouter for the llm layer vercel ai sdk for streaming. but for agentic stuff specifically skip langchain. anthropic's agent sdk or just raw tool calling loops with structured outputs. less abstraction = easier debugging when agents go sideways"
X Link 2026-02-06T18:35Z [---] followers, [--] engagements

"@rahulgs smart move from openai tbh. keeping codex app-only means the narrative stays "try it yourself" instead of getting reduced to a benchmark number. but also means devs building on the api are stuck with models they can actually test independently"
X Link 2026-02-06T20:40Z [---] followers, [---] engagements

"@tannerlinsley @opencode which part of the UX keeps pulling you back the vim-style keybinds or something else"
X Link 2026-02-06T22:12Z [---] followers, [---] engagements

"@tannerlinsley @opencode the line selection thing is a nice touch. claude code TUI feels more functional than polished rn which is fine for terminal devs but opencode is clearly investing in the DX layer"
X Link 2026-02-06T23:16Z [---] followers, [---] engagements

"@natolambert convergence happening faster than anyone expected. both heading toward the same place just from different starting points. codex starting from model strength claude code from workflow integration"
X Link 2026-02-06T23:26Z [---] followers, [---] engagements

"@aliByteCode honestly the [--] person thing is less true now than even [--] months ago. claude code handles my dev design iterations even first drafts of marketing copy. bottleneck shifted from "doing everything" to "deciding what to do next" which is harder tbh"
X Link 2026-02-06T23:33Z [---] followers, [--] engagements

"@gregisenberg the "disappears for a while" part is undersold imo. ran opus agent teams on a refactor yesterday and the surprising thing isnt speed its that it makes decisions you wouldnt have made. sometimes better sometimes worse. did you notice that in the polymarket build"
X Link 2026-02-06T23:40Z [---] followers, [---] engagements

"@daniel_mac8 already doing this. opus for the interactive stuff codex running background tasks simultaneously. no orchestrator needed just two terminals"
X Link 2026-02-07T01:19Z [---] followers, [---] engagements

"the routing layer is tempting but honestly ive found the overhead isnt worth it yet. right now i just mentally route: anything that needs back and forth goes to opus anything i can describe once and walk away goes to codex. the model strengths make the routing obvious enough that automating it adds complexity without much gain https://twitter.com/i/web/status/2019966102454112610 https://twitter.com/i/web/status/2019966102454112610"
X Link 2026-02-07T02:47Z [---] followers, [---] engagements

"@JoeAbbud @daniel_mac8 oh thats clever. shared MCP as the memory layer between them. hadnt tried that angle yet"
X Link 2026-02-07T04:28Z [---] followers, [--] engagements

"@ns123abc knowing what good code looks like is exactly what makes claude code useful tho. the people getting the most out of it are the ones who can tell when its wrong in [--] seconds instead of debugging for an hour"
X Link 2026-02-07T04:35Z [---] followers, [---] engagements

"@hey_atticus @daniel_mac8 opposite actually. opus for frontend because it has better design judgment and component composition. codex i run on backend stuff migrations scripts. the visual stuff needs more taste than speed"
X Link 2026-02-07T07:30Z [---] followers, [--] engagements

"lol same. updated my project config [--] times last month just to keep up with new capabilities. stopped recommending specific setups started recommending the meta pattern: one index file pointing to knowledge dirs skills for repeatable workflows. tools change but that structure survives every switch https://twitter.com/i/web/status/2020039524752203797 https://twitter.com/i/web/status/2020039524752203797"
X Link 2026-02-07T07:38Z [---] followers, [---] engagements

"@VadimStrizheus stack looks right but that last 10% claude doesnt write is where all the debugging time goes. same setup here minus lovable add convex instead of supabase for the reactive queries. biggest bottleneck isnt code generation tho its knowing when the output is wrong"
X Link 2026-02-07T07:40Z [---] followers, [--] engagements

"@levie the gap isnt just attention tho its iteration speed. people who ship and test with these tools compound way faster than people who just learn them. seen devs who know every model benchmark but havent shipped anything vs founders quietly 10xing output"
X Link 2026-02-07T07:42Z [---] followers, [---] engagements

"@xlr8harder lol yeah this is the one thing that drives me back to claude code every time. codex defaults to 60s timeouts on everything even npm install on a fresh project"
X Link 2026-02-07T07:49Z [---] followers, [--] engagements

"@robinebers [---] daily users and climbing thats a nice ramp. the windsurf plugin angle is smart too meeting devs where they already are instead of asking them to install something new"
X Link 2026-02-07T10:45Z [---] followers, [--] engagements

"@camsoft2000 the 2x codebase problem isnt writing the code its maintaining state parity when your backend changes. AI can scaffold both native apps fast but the moment you need to debug a platform specific edge case on iOS vs Android you still need someone who knows what theyre looking at"
X Link 2026-02-07T11:06Z [---] followers, [---] engagements

"@thepushkarp @claudeai @OpenAI i use both but for different things. codex is great for async tasks you can fire and forget but claude code still wins for the interactive sessions where you need back and forth. the timeout defaults on codex (60s) get annoying fast when youre debugging something complex"
X Link 2026-02-07T13:18Z [---] followers, [--] engagements

"honestly i think the editor IS the wrong abstraction for this. the terminal + task runner pattern works better. i run parallel claude code agents with file ownership scoping and its way cleaner than trying to manage it all visually. the future isnt a better editor its better orchestration https://twitter.com/i/web/status/2020126488545358261 https://twitter.com/i/web/status/2020126488545358261"
X Link 2026-02-07T13:24Z [---] followers, [--] engagements

"@TheOvermanEthos worked at bosch for [--] years before quitting to start my own thing. the smartest engineers i knew there are still there [--] years later making great salaries. not because they lack courage because the golden handcuffs get heavier every year. each promotion makes leaving harder"
X Link 2026-02-07T13:26Z [---] followers, [--] engagements

"@melvynxdev lol welcome to the club. running [--] claude code sessions in parallel tmux panes is peak "i have no idea whats happening but its working""
X Link 2026-02-07T13:30Z [---] followers, [---] engagements

"@robinebers the uglier the UI the deeper the moat lol. means they won on distribution not design"
X Link 2026-02-07T16:13Z [---] followers, [--] engagements

"@petergyang @kieranklaassen the skill running for hours is the real flex. we do something similar with claude code skills but never thought about recording demos as part of the output. thats clever. the plan-build-test loop maps well to how skills work when you add grading rubrics for self-checking"
X Link 2026-02-07T16:22Z [---] followers, [---] engagements

"@aarondfrancis been using tmux + claude code but the session management is painful. does solo handle the context handoff between panes or is it mostly visual"
X Link 2026-02-07T18:00Z [---] followers, [---] engagements

"@gracepace_ honestly most wrappers i use daily are the boring ones. perplexity for search cursor for when i need quick edits v0 for UI scaffolding. the ones doing well arent doing anything fancy they just nail one specific workflow"
X Link 2026-02-07T19:21Z [---] followers, [--] engagements

"@ryan_t_brown @Codie_Sanchez thats the move. agency gives you the customer conversations that tell you what to build next. most saas founders skip that step and guess"
X Link 2026-02-07T19:26Z [---] followers, [--] engagements

"been testing it for the last hour and the speed difference is noticeable but what im really wondering about is whether fast mode changes how you structure prompts. like with regular opus i tend to front-load context because i know itll process it all. does faster output change how you think about input https://twitter.com/i/web/status/2020218436551594051 https://twitter.com/i/web/status/2020218436551594051"
X Link 2026-02-07T19:29Z [---] followers, [----] engagements

"@alexcooldev the distinction isnt vibe vs blind tho. its prototype vs product. vibe coding gets you to prototype fast. shipping a product someone pays for monthly still requires understanding the boring parts"
X Link 2026-02-07T21:21Z [---] followers, [--] engagements

"@artee_49 codex bundles usage into the chatgpt plus sub so the limits feel invisible. claude code burns through opus tokens fast on max plan too especially with agent teams. which plan are you on"
X Link 2026-02-07T23:01Z [---] followers, [--] engagements

"@dkubb @camsoft2000 yeah thats basically the electron/tauri pattern. works until the shims need platform-specific UX then youre writing native anyway. depends on how thin you can keep that layer"
X Link 2026-02-07T23:06Z [---] followers, [--] engagements

"salesforce bought heroku for the developer ecosystem then spent a decade trying to make it fit the enterprise CRM playbook. same pattern every time: big co acquires dev tool strips what made it good wonders why devs leave. meanwhile hetzner charges me 40/month for what heroku wanted $250 for https://twitter.com/i/web/status/2020274270715470303 https://twitter.com/i/web/status/2020274270715470303"
X Link 2026-02-07T23:11Z [---] followers, [---] engagements

"the speed pricing makes sense when you think about it from the workflow side. been running opus in claude code for months and the bottleneck isnt the model thinking its me waiting for output on long refactors. if fast mode cuts a [--] second agent run to [--] seconds thats not 6x cost thats 6x more iterations per hour. compounding effect on shipping speed https://twitter.com/i/web/status/2020276475753345396 https://twitter.com/i/web/status/2020276475753345396"
X Link 2026-02-07T23:20Z [---] followers, [----] engagements

"@AISpout @theo 4o is still surprisingly good for quick stuff. but for anything multi-file opus just gets it done with less back and forth"
X Link 2026-02-08T01:19Z [---] followers, [---] engagements

"@martin_casado @cursor_ai @convex single-writer per chunk is the right call. convex reactive queries love that pattern. are you planning to open up multi-editor on shared maps later or does the portal model keep things clean enough"
X Link 2026-02-08T01:20Z [---] followers, [--] engagements

"@dkubb @camsoft2000 yeah its a spectrum. apps like slack where 90% is ui logic shims everywhere. something like a video encoder where the core is heavy shared rust core makes perfect sense"
X Link 2026-02-08T01:21Z [---] followers, [--] engagements

"rls is painful yeah lol. we keep policies dead simple per table and push complex logic to edge functions instead. trying to do multi-join authorization in rls policies is where it gets ugly fast. ai actually helps write the policies but you still gotta test every role manually or youll ship a hole https://twitter.com/i/web/status/2020472247086141896 https://twitter.com/i/web/status/2020472247086141896"
X Link 2026-02-08T12:18Z [---] followers, [--] engagements

""vibe coding" had a good run but in [--] months nobody will say it itll just be coding like how we stopped saying "agile" every sentence in 2015"
X Link 2026-02-08T15:00Z [---] followers, [--] engagements

"so i built a tool in like [--] days with claude + supabase. total infra cost maybe $10/mo. then tried to get users and spent $200 on twitter ads got [--] clicks [--] signups. the building part is democratized now but distribution is still the hard part. best ROI was honestly just posting about it here and replying to people in the niche https://twitter.com/i/web/status/2020720376997982307 https://twitter.com/i/web/status/2020720376997982307"
X Link 2026-02-09T04:44Z [---] followers, [--] engagements

"already seeing this play out in real time. we moved all our infra from AWS to hetzner (german servers) last year after watching the cloud pricing games. total cost went from $400/mo to $45/mo for the same workloads. the sovereign stack thing isnt just geopolitics its economics too https://twitter.com/i/web/status/2020720754510479754 https://twitter.com/i/web/status/2020720754510479754"
X Link 2026-02-09T04:45Z [---] followers, [----] engagements

"@zarazhangrui the plumbing part is what nobody warns you about. spent more time connecting stripe webhooks to supabase than building actual features on my last project"
X Link 2026-02-09T08:01Z [---] followers, [---] engagements

"vibe coding and agentic engineering arent stages theyre different skills. vibe coding is product thinking with ai execution. agentic engineering is understanding what the ai does when it breaks. karpathy is right that the shift is happening but most teams need both on the same person now https://twitter.com/i/web/status/2020771246926147624 https://twitter.com/i/web/status/2020771246926147624"
X Link 2026-02-09T08:06Z [---] followers, [--] engagements

"@_baretto that growth curve from mid-2024 onward tho. 100s of micro improvements compounding is exactly how bootstrapped hits [--] figures"
X Link 2026-02-09T11:15Z [---] followers, [--] engagements

"@solarise_webdev @johnrushx laravel on the entry level box is a good stress test too. if it handles that well at low traffic scaling up later is just bumping the plan"
X Link 2026-02-09T11:30Z [---] followers, [--] engagements

"@raindrop_io oauth [---] is a smart call. most mcp servers ive tried skip auth entirely or roll their own janky token flow. hows the refresh work with streaming connections tho"
X Link 2026-02-09T13:16Z [---] followers, [---] engagements

"@mattpocockuk been building with ai tools daily for a year and the compounding shows. each workflow you figure out stacks on the last one. devs who start now vs devs who start in [--] years will have completely different skill ceilings"
X Link 2026-02-09T16:31Z [---] followers, [----] engagements

"@antirez been running both daily. speed debate is wrong framing imo. opus for interactive sessions where you course correct in real time codex for background tasks you check an hour later. different loops not competing tools"
X Link 2026-02-09T20:56Z [---] followers, [---] engagements

"@jessethanley @Hetzner_Online @Cloudflare running [--] sites on hetzner + cloudflare rn. total bill is under [--] euros. vercel pricing only makes sense if you genuinely need edge compute"
X Link 2026-02-09T22:15Z [---] followers, [----] engagements

"@ChaiKamNamak @emollick yeah fair i defaulted to coding because thats what i test against daily. whats the study actually measuring"
X Link 2026-02-09T23:07Z [---] followers, [--] engagements

"@udiWertheimer tool knowledge fades yeah but workflow patterns dont. my project configs and grading rubrics work across claude code opencode whatever ships next month. the meta skill is knowing how to set up the harness not memorizing which model flag does what"
X Link 2026-02-09T23:16Z [---] followers, [---] engagements

"the gap isnt even between trained vs untrained anymore. its between people who know what the output should look like vs people who cant tell when its wrong. ran [--] projects across claude code and the biggest differentiator is whether i can spot a bad architecture decision in the first [--] seconds of output https://twitter.com/i/web/status/2021040042278519120 https://twitter.com/i/web/status/2021040042278519120"
X Link 2026-02-10T01:54Z [---] followers, [--] engagements

""claude code keeps asking for confirmation thats so annoying" bro thats a feature you want an ai agent that just runs random code without asking the security review is the product not the bug"
X Link 2026-02-10T03:00Z [---] followers, [---] engagements

"@aakashgupta using both daily and the practical worry isnt model quality degrading. its the UX shifting from tasks-completed to time-spent. every ad-funded product eventually optimizes for keeping you in the app longer instead of getting you out faster"
X Link 2026-02-10T05:42Z [---] followers, [---] engagements

"@tdinh_me at $46k/month you could literally hire a devops person to manage self hosted infra and still save money lol"
X Link 2026-02-10T05:52Z [---] followers, [---] engagements

"@ALEngineered honest question tho what would the ideal interview look like every time someone says "just test if they can build" the interview ends up being. build something in [--] hours under pressure which isnt much better"
X Link 2026-02-10T08:07Z [---] followers, [---] engagements

"@levifig @tannerlinsley oh you cant just tease that and not share lol. whats the project"
X Link 2026-02-10T13:27Z [---] followers, [--] engagements

"@boxmining the spending adds up fast but its cheaper than the debugging hours it saves. real split for me is opus for anything interactive where judgment matters codex for background tasks i dont want to babysit. composer i keep trying but the extensibility gap keeps pulling me back"
X Link 2026-02-10T13:44Z [---] followers, [--] engagements

"@eliana_jordan tale as old as open source. wrapper wave hits 90% die in [--] months the 10% that add real DX value stick around"
X Link 2026-02-10T13:50Z [---] followers, [--] engagements

"@HsanC_ guilty. stripe tab has been open since january"
X Link 2026-02-10T17:07Z [---] followers, [--] engagements

"@vikashkprajapat @thepatwalls what kind of youtube tools always looking for stuff that saves time on the production side"
X Link 2026-02-11T01:45Z [---] followers, [--] engagements

"@damianplayer best framing ive seen on this. the people panicking about AI are usually the last ones to actually try using it"
X Link 2026-02-11T10:28Z [---] followers, [--] engagements

"@ExpenseLens haha the premature feature phase. most dangerous stage because it feels productive"
X Link 2026-02-11T21:14Z [---] followers, [--] engagements

"@ExpenseLens thats the best part. real feedback analytics dashboards"
X Link 2026-02-11T21:42Z [---] followers, [--] engagements

"@ExpenseLens lol better opus shreds it now than users shred it publicly"
X Link 2026-02-11T22:36Z [---] followers, [--] engagements

"@jackfriks route through openrouter. cheaper models for scaffolding steps frontier model only for final output. openclaw loops burn tokens way faster than you expect"
X Link 2026-02-12T01:40Z [---] followers, [---] engagements

"if youre building a one-time purchase app: jetbrains model is the answer buy once + [--] year of updates after that pay for new features or keep what you have everyone else is trying to shove subscriptions where they dont fit"
X Link 2026-02-12T03:00Z [---] followers, [--] engagements

"yeah the harness thesis keeps proving itself. been running self-check loops where claude validates its own api responses against expected schemas before moving to the next feature. catches 80% of the mock data drift before it compounds. model doesnt matter if your testing layer is thin https://twitter.com/i/web/status/2020966036007420222 https://twitter.com/i/web/status/2020966036007420222"
X Link 2026-02-09T21:00Z [---] followers, [---] engagements

"@seloesque moved to hetzner + coolify a while back and genuinely forgot what a hosting bill looks like. vercel pricing model makes zero sense once you outgrow the free tier"
X Link 2026-02-10T08:12Z [---] followers, [----] engagements

"@Yuchenj_UW the disruption isnt replacing slack. its 10M people who stop paying $50/month for generic saas because they can build exactly what they need themselves. thats a different market collapse"
X Link 2026-02-10T15:16Z [---] followers, [---] engagements

"@alexcooldev whats driving more of that web traffic your organic tiktok or the ugc stuff"
X Link 2026-02-10T16:58Z [---] followers, [---] engagements

"@JRupena schwab doesnt need better ai tho. they have distribution and regulatory moats that took decades. altruists real competition is every other fintech startup trying the same ai-native play"
X Link 2026-02-10T17:08Z [---] followers, [---] engagements

"@expo @claudeai @cursor_ai this is where mcp clicks for real. managing builds from inside your coding agent instead of switching to a dashboard. ran a similar setup with supabase mcp for db ops and the context-switching reduction alone saved me 20+ minutes per session"
X Link 2026-02-10T18:49Z [---] followers, [---] engagements

"@nummanali @cursor_ai whats the agentic depth like compared to codex speed is great but my bottleneck is usually multi-file refactors not single completions"
X Link 2026-02-10T18:51Z [---] followers, [---] engagements

"@garrytan translation tasks like this are where AI absolutely crushes it tho. clear input clear output 1:1 mapping. the harder vibe coding test is building something new where the requirements are still half-formed"
X Link 2026-02-10T18:57Z [---] followers, [---] engagements

"@flaviocopes and then vibe coding showed up and mass produced it lol"
X Link 2026-02-10T20:04Z [---] followers, [--] engagements

"@iruletheworldmo running both. codex for fire and forget background tasks claude code when i need to steer multi-file changes interactively. theyre different tools at this point not competitors"
X Link 2026-02-10T22:22Z [---] followers, [---] engagements

"@kasterkeqi @qasimbizs depends what you mean by self host. moved off vercel to hetzner ages ago (4/month lol). but supabase the auth + rls + realtime stack is worth the price even at scale. self hosting postgres and building all that from scratch is where the real cost hides"
X Link 2026-02-10T22:28Z [---] followers, [---] engagements

"@simonw finally. 'it passes tests' means nothing when the agent rewrote your entire auth flow to make one test green lol"
X Link 2026-02-10T22:34Z [---] followers, [--] engagements

"@expo @claudeai @cursor_ai edtech platform that indexes 60k+ vocational jobs in germany. planning a mobile companion app for job alerts and employer matching. the mcp server is honestly what pushed expo up my shortlist"
X Link 2026-02-10T23:52Z [---] followers, [--] engagements

"@ImSh4yy the signal isnt the hype its what people are still using [--] weeks later"
X Link 2026-02-10T23:59Z [---] followers, [--] engagements

"the "same label completely different financial instrument" part is the thing. been bootstrapping for [--] years and watched competitors raise 10x what we ever made in revenue hire [--] people then quietly shut down while we kept growing. the normalization isnt just about round sizes its about what founders optimize for. when your seed is $60M you optimize for the next raise not for customers. https://twitter.com/i/web/status/2021374894760657228 https://twitter.com/i/web/status/2021374894760657228"
X Link 2026-02-11T00:05Z [---] followers, [--] engagements

"@kirodotdev right call. been mixing models for months and the difference is huge. opus for anything creative or frontend gpt for backend logic smaller ones for linting and quick fixes. the "one model for everything" default is leaving so much on the table"
X Link 2026-02-11T00:12Z [---] followers, [---] engagements

"@RONYNWA10 @JRupena for retail yeah totally robinhood UX is miles ahead. but altruist targets advisors not retail. different switching cost calculus firms move way slower than individual accounts"
X Link 2026-02-11T01:48Z [---] followers, [--] engagements

"@AnthropicAI 0.83% to 0.04% refusal rate while actually raising the safety bar. thats the part most people skim past. separating safety from over-caution is way harder than just making the model smarter"
X Link 2026-02-11T01:50Z [---] followers, 12K engagements

"@KabirGoel the company claude leaderboard is the new eng ladder"
X Link 2026-02-11T02:01Z [---] followers, [---] engagements

"@Yuchenj_UW seen this play out before. research cofounders bounce when the company shifts from "figure out what works" to "ship what works." different phases different people. doesnt mean the ship is sinking"
X Link 2026-02-11T02:03Z [---] followers, [----] engagements

"@emollick biggest shift ive noticed too. the useful updates now are all about integrations not benchmarks. nobody cares about a 3% mmlu bump but claude inside powerpoint changes how you actually work"
X Link 2026-02-11T02:16Z [---] followers, [---] engagements

"@arvidkahl the review part is what people skip over. ai output looks correct to beginners. experienced devs spot the architecture smell in seconds and that gap just keeps compounding"
X Link 2026-02-11T06:46Z [---] followers, [--] engagements

"@chainyoda theres a gap between vibe coding and using AI tooling properly tho. nobody should be one-shotting an ERP. but running claude code with specs tests and code review thats just faster engineering. the hate only makes sense if you assume all AI coding means zero oversight"
X Link 2026-02-11T22:52Z [---] followers, [--] engagements

"@chongdashu yeah "does this feel fun" is basically impossible to write a test for lol"
X Link 2026-02-11T23:47Z [---] followers, [--] engagements

"@BphilSoChill @harjotsgill skills yeah those are just markdown port in seconds. the hooks and mcp configs are where switching cost actually lives. took months to get those right"
X Link 2026-02-12T13:21Z [---] followers, [--] engagements

"@BphilSoChill @harjotsgill fair enough. if codex adds hooks later itll be basically the same setup. convergence is happening fast"
X Link 2026-02-12T15:00Z [---] followers, [--] engagements

"@rahim_mirani been at that top tier for [--] years except from india not SF. no visa hurdles but cross-border payments and entity structuring across [--] countries adds a whole different layer. skipping the raise part was the best decision tho"
X Link 2026-02-12T23:26Z [---] followers, [--] engagements

"@steipete $80M for a wrapper that 504s on launch day lmao. you shipped the actual thing for free"
X Link 2026-02-09T18:57Z [---] followers, [----] engagements

"@aramh the types part tracks. any constraint you give the model upfront saves 3x debugging downstream. but jumping from "types are useful" to "pure functional or nothing" is doing heavy lifting. typescript strict mode catches 90% of what matters in practice without the haskell overhead"
X Link 2026-02-10T13:39Z [---] followers, [---] engagements

"been saying this for a while. coding was never the bottleneck for most failed startups. distribution was. and now with vibe coding you can build faster but you also burn through bad ideas faster without learning why they failed. ran multiple products over [--] years and the ones that worked had nothing to do with technical complexity. they worked because the market was there before we wrote a single line. https://twitter.com/i/web/status/2021379169289109718 https://twitter.com/i/web/status/2021379169289109718"
X Link 2026-02-11T00:22Z [---] followers, [---] engagements

"hot take: solo founders should go b2b b2c kills you with: - marketing budget you dont have - support volume you cant handle - churn you cant outrun b2b [--] good customers = sustainable business did both. b2b every time"
X Link 2026-02-11T03:00Z [---] followers, [---] engagements

"@lovish888 @taimurabdaal lol nah claude code is the one that stuck. tried everything else but kept coming back. havent looked at antigravity tho you using it"
X Link 2026-02-11T10:22Z [---] followers, [--] engagements

"@heygeorgekal same on the credit card thing. but the one that actually makes me close the tab faster is "contact sales for pricing". if i have to get on a call to learn what it costs im out"
X Link 2026-02-11T10:29Z [---] followers, [--] engagements

"best story about sarvam: a menlo ventures partner called their first model launch "embarrassing." [--] downloads in [--] days. [--] months later same VC: "i cant remember the last time an indian software product made such a strong impression." from embarrassing to impressive. thats how you build"
X Link 2026-02-11T13:39Z [---] followers, [--] engagements

"the indian government picked sarvam to build the countrys sovereign LLM under the IndiaAI mission. [--] billion parameters. multimodal. built and deployed entirely in india. khosla ventures lightspeed peak XV backed them with $54M. india is betting on itself for AI infrastructure. https://twitter.com/i/web/status/2021579800222089505 https://twitter.com/i/web/status/2021579800222089505"
X Link 2026-02-11T13:39Z [---] followers, [--] engagements

"@lucas_montano yeah this one hit me too. used to be muscle memory ctrl+c to clear and retype. now it kills whatever claude is doing mid-run and i have to start over"
X Link 2026-02-11T15:23Z [---] followers, [--] engagements

"the delegation gap tracks. running multiple products taught me the bottleneck isnt the code generation its the decision layer between tasks. which agent runs next what context to pass forward when to override. [--] min of orchestration saving [--] hours of execution but you cant automate the orchestration itself yet https://twitter.com/i/web/status/2021890822841782432 https://twitter.com/i/web/status/2021890822841782432"
X Link 2026-02-12T10:15Z [---] followers, [--] engagements

"@heygeorgekal coolify. self hosted vercel alternative on a hetzner box. deploy unlimited projects zero egress fees and you actually own the infra"
X Link 2026-02-12T13:46Z [---] followers, [--] engagements

"@DataChaz @TypingMindApp @t3dotchat @AnythingLLM same split here. claude for code gpt for backend logic gemini for anything visual. the ui problem isnt going away tho half these tools have great models behind terrible interfaces"
X Link 2026-02-12T15:12Z [---] followers, [--] engagements

"@2sush the pen was never the test. they just wanted to see if youd freeze. which. we all did"
X Link 2026-02-12T18:11Z [---] followers, [--] engagements

"@47max7 @antirez fair hooks and the skills bash layer are whats keeping me on claude code specifically. the rest has converged yeah"
X Link 2026-02-12T23:07Z [---] followers, [--] engagements

"@tristanbob debugging supabase from claude code right now means tab-switching to the dashboard to grab logs manually. a CLI logs command would close that loop completely"
X Link 2026-02-13T00:51Z [---] followers, [--] engagements

"@hnshah been learning this the hard way with ai coding. spent [--] days fighting claude on a migration script because my initial prompt was vague. rewrote the prompt with exact schema details and column mappings got it in one shot. the prompt IS the product at this point"
X Link 2026-02-13T07:25Z [---] followers, [--] engagements

"@nummanali ran into the same compaction amnesia with claude code. my fix was dumping the plan into a markdown file at project root so every new context loads it automatically. works way better than keeping it in the agents head"
X Link 2026-02-13T17:02Z [---] followers, [---] engagements

"@fchollet same thing with wireframes. ended up using mermaid through claude code for anything structural. image gen for creative stuff code gen for precision. splitting them that way saved me a ton of time"
X Link 2026-02-13T22:16Z [---] followers, [---] engagements

"@DanielMiessler @theo @pedramamini been using the ralph wiggum technique for autonomous builds for a while. the git worktrees part is what most people skip tho isolated branches per run means you review clean diffs instead of untangling one messy commit history"
X Link 2026-01-31T19:42Z [---] followers, [---] engagements

"@Akshat_World did exactly this. moved to germany invest in US/EU etfs. the mental relief of not watching every budget announcement for surprise taxes is worth the move alone"
X Link 2026-02-02T06:11Z [---] followers, 141.9K engagements

"compounded loops are the one i keep coming back to. been running ralph wiggum loops overnight where it iterates until tests pass. trick is designing good exit conditions so it converges instead of spinning forever. parallel windows are great but i keep losing track of which agent is doing what https://twitter.com/i/web/status/2019456224086094244 https://twitter.com/i/web/status/2019456224086094244"
X Link 2026-02-05T17:00Z [---] followers, [--] engagements

"@supabase @claudeai been using the supabase mcp server in claude code for weeks now. running queries and migrations without leaving the terminal is genuinely faster than the dashboard for most things"
X Link 2026-02-05T19:58Z [---] followers, [----] engagements

"@steipete @_zenzoen lived in munich for years. this is peak german energy lol"
X Link 2026-02-11T16:56Z [---] followers, [---] engagements

"@FI_InvestIndia hospitals literally run two rate cards. cash vs insurance. been like this for years and nobody talks about it"
X Link 2026-02-11T16:57Z [---] followers, [---] engagements

"@matic_branko @SirPromptwright lovable bolt and v0 for the no-code-ish side. the stuff i see non-tech people ship works for internal tools and mvps. youre right that most of it falls apart at scale tho. the ones that survive usually bring in someone technical eventually"
X Link 2026-02-11T18:19Z [---] followers, [---] engagements

"@fortelabs been on terminal-first tools daily and that feeling never fades. something about a blinking cursor makes you think harder about what youre asking for. guis abstract that away"
X Link 2026-02-11T18:23Z [---] followers, [--] engagements

"@matic_branko @SirPromptwright yeah fair enough. the shipping part works but the maintaining part is where it falls apart without eng. the narrative oversells it"
X Link 2026-02-11T19:42Z [---] followers, [---] engagements

"@simonw using claude to test openai features before writing them up might be the most [----] thing ive read today"
X Link 2026-02-11T19:44Z [---] followers, [---] engagements

"@emollick seeing this in pricing already. opus fast mode costs 6x more for 2.5x speed. thats compute scarcity priced in. and once claude ships inside powerpoint and excel the enterprise token burn alone is going to be staggering"
X Link 2026-02-11T19:45Z [---] followers, [---] engagements

"@richiemcilroy @cap "every bug report cuts deep" after [---] years lol. chrome extension alongside native is a smart play gives users a fallback while you nail the last 5%"
X Link 2026-02-11T19:53Z [---] followers, [--] engagements

"@litti_chokha0 @SirPromptwright scroll this timeline for [--] min and youll find at least [--] people posting mrr updates who started coding [--] months ago with ai tools. theyre not building the next figma but landing pages chrome extensions small automation tools are all within reach now"
X Link 2026-02-11T20:04Z [---] followers, [---] engagements

"@theo what'd you test it on been burning through tokens on agentic scaffolding and need a cheaper model that doesnt fall apart on tool calls"
X Link 2026-02-11T21:38Z [---] followers, [----] engagements

"@nummanali @claudeai the "numman returns" welcome is a nice touch. getting the whole team on claude at once is the move way better than one person trying to sell it internally"
X Link 2026-02-11T21:40Z [---] followers, [--] engagements

"@burkov the "switching" framing is off tho. most devs i know run both depending on the task. opus for frontend codex for backend same session. the moat isnt the model its the workflow built around it"
X Link 2026-02-11T22:43Z [---] followers, [---] engagements

"@brungarc sentry buying MCP tooling while zeeg is testing claude cowork for corp dev. the signal is clear devtools companies are going all in on the AI agent layer. MCP is becoming the integration standard whether you planned for it or not"
X Link 2026-02-11T22:58Z [---] followers, [--] engagements

"@trikcode trust fall with your api keys"
X Link 2026-02-12T15:08Z [---] followers, [---] engagements

"@thsottiaux @TeksEdge @OpenAI does it need to be more agentic or just better at staying on task every time i give codex a multi-step refactor it loses the plot by step 3"
X Link 2026-02-12T16:45Z [---] followers, [---] engagements

"@joshuawhocodes @adamwathan @stripe true on desktop. but try buying something in mobile safari on a friends phone lol. thats where link picks up the slack"
X Link 2026-02-12T21:48Z [---] followers, [--] engagements

"@batuhan lol the tradeoff is worth it for most tasks. i keep opus around for the stuff that actually needs thinking"
X Link 2026-02-13T00:45Z [---] followers, [---] engagements

"@FeifanZ eh every hosted service goes down eventually. postgres under the hood means you can always migrate if it gets bad. the DX tradeoff is still worth it for most projects imo"
X Link 2026-02-13T02:38Z [---] followers, [--] engagements

"@heygeorgekal pricing always. "book a demo" is code for "we havent figured out pricing yet" or "its expensive". either way i bounce. even ballpark ranges help people self-qualify"
X Link 2026-02-13T09:56Z [---] followers, [--] engagements

"type [--] is what actually happens to 80% of companies: bought [--] agent subscriptions one person uses it rest forgot the login. the gap isnt between agent-first and agent-resistant. its between companies that actually integrate ai into workflows vs ones that just bought the tools and called it a strategy https://twitter.com/i/web/status/2022253453008146887 https://twitter.com/i/web/status/2022253453008146887"
X Link 2026-02-13T10:16Z [---] followers, [--] engagements

"@ZssBecker give it [--] months and someones gonna post a linkedin carousel titled 'meet my team' featuring [--] claude instances and a cron job"
X Link 2026-02-13T14:04Z [---] followers, [---] engagements

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