@udaysy Uday Yatnalli

Uday Yatnalli posts on X about ai, claude code, code, vibe coding the most. They currently have [---] followers and [---] posts still getting attention that total [------] engagements in the last [--] hours.

Engagements: [------] #

• [--] Week [-------] +75%
• [--] Month [-------] +1,709%
• [--] Months [-------] +2,720,052%

Mentions: [--] #

• [--] Week [---] +6.40%
• [--] Month [---] +176%
• [--] Months [---] +46,100%

Followers: [---] #

• [--] Week [---] +16%
• [--] Month [---] +185%
• [--] Months [---] +1,559%

CreatorRank: [-------] #

Social Influence

Social category influence technology brands 22.76% finance 4.48% social networks 4.1% countries 3.73% stocks 2.99% vc firms 0.37% products 0.37% travel destinations 0.37%

Social topic influence ai 14.18%, claude code #403, code 6.72%, vibe coding #520, lol 2.99%, agentic #1049, products #1901, openclaw #3485, loops 2.61%, open ai #978

Top accounts mentioned or mentioned by @antirez @theo @robinebers @cristianuibar @steipete @emollick @heygeorgekal @fortelabs @expenselens @openclaw @yuchenjuw @claudeai @cursorai @nummanali @simonw @sirpromptwright @bphilsochill @theahmadosman @joshuajohnsonai @giorgiottolina

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

Top Social Posts

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

"@cristianuibar havent tried grok for anything serious yet. the live data angle is cool but im too deep in the claude + gpt workflow rn"
X Link 2026-02-14T23:14Z [---] followers, [--] engagements

"watched someone automate their entire suno workflow: - claude code generates jingle lyrics - chrome extension opens suno - fills in lyrics style title - hits create - does this across [--] tabs in parallel no api. no code. just browser automation"
X Link 2026-01-13T08:31Z [---] followers, [--] engagements

"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

"@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

"@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

"@Yuchenj_UW the sandbox limitation alone disqualifies codex for anything ml-heavy. claude code runs on your actual machine with your actual gpu. that gap matters more than any model benchmark"
X Link 2026-02-04T21:29Z [---] followers, [----] engagements

"that dashboard tells the whole story lol. 527K function calls 600MB storage everything basically idle but bandwidth at 3x overage. feels like theyre billing for reactive query syncs that most apps dont even need at that volume. ran into the same thing one bad subscription query was responsible for like 80% of my bandwidth https://twitter.com/i/web/status/2019454163579728364 https://twitter.com/i/web/status/2019454163579728364"
X Link 2026-02-05T16:52Z [---] followers, [--] engagements

"context engineering is exactly the right framing. been running [--] projects with files and the biggest lesson is treating it like an index not a dump. keep the main file under [---] lines pointing to knowledge/ directories for deep context. without that structure it rots fast and the model starts ignoring instructions halfway through http://CLAUDE.md http://CLAUDE.md"
X Link 2026-02-05T16:54Z [---] 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

"@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

"@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

"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

"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

"@iannuttall wait till they add the region suffix. gpt-5.3-codex-xhigh-xfast-us-east-2"
X Link 2026-02-09T11:26Z [---] 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

"@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

"@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

"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

"@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

"@alexcooldev the gamification is the easy part tho. duolingo's actual edge is like [--] years of A/B testing notification timing and streak psychology. the clones always copy the surface and skip the hard stuff"
X Link 2026-02-10T08:00Z [---] 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

"@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

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

"@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

"@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

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

"@trikcode lol the idea validation loop just got 100x faster. ship it get [--] users realize nobody wanted it. used to take [--] months now takes a weekend"
X Link 2026-02-10T17:12Z [---] 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

"the 10:1 ratio makes sense but the architect role is 10x harder to fill than the PM roles. review is the bottleneck now. someone who can look at ai-generated code and know when its subtly wrong. thats a different skill than building from scratch. already seeing this with small teams. scope per person expanded but hiring for the verification layer is brutal https://twitter.com/i/web/status/2021311018849403155 https://twitter.com/i/web/status/2021311018849403155"
X Link 2026-02-10T19:51Z [---] followers, [----] engagements

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

"@theo youtube + cheap components + zero fear of breaking things. this generation of builders is gonna be something else"
X Link 2026-02-10T21:13Z [---] followers, 12K 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

"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

"@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

"@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

"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

"@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

"@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

"@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

"@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

"switched from supabase to convex for a new project supabase workflow: claude generates migration file i review [--] lines of sql deploy something breaks rollback debug for [--] mins convex workflow: claude edits schema.ts deploys done fewer files means fewer places for ai to introduce subtle bugs https://twitter.com/i/web/status/2021600112720437305 https://twitter.com/i/web/status/2021600112720437305"
X Link 2026-02-11T15:00Z [---] followers, [---] engagements

"@kaif9999 zomato and swiggy combined costing 3x your entire AI stack is peak indian dev culture. we optimize cloud costs to save $5/mo then order biryani at 2am without thinking twice"
X Link 2026-02-11T15:13Z [---] 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

"@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

"the [---] docs part is what makes this click. same pattern i see with claude code more context = better output. biggest shift for me was going from "generate this one thing" to "heres my entire project figure out what needs fixing". cowork looks like that leap but for knowledge work https://twitter.com/i/web/status/2021654028896338001 https://twitter.com/i/web/status/2021654028896338001"
X Link 2026-02-11T18:34Z [---] 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

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

"smart infrastructure play disguised as goodwill. paying grid costs upfront = faster permitting = faster capacity buildout. while everyone else fights NIMBYism and utility pushback anthropic is basically buying the fast lane. the companies that solve power constraints first win the next [--] years of AI scaling https://twitter.com/i/web/status/2021697042779975835 https://twitter.com/i/web/status/2021697042779975835"
X Link 2026-02-11T21:25Z [---] 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

"@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

"@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

"@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

"@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

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

"@lennysan @openclaw biggest lesson: route through openrouter instead of direct api keys. agentic loops burn tokens 5-10x faster than youd expect and per-minute caps kill your sessions before context does. use cheaper models for scaffolding steps and frontier only for the final output"
X Link 2026-02-11T23:51Z [---] followers, [---] engagements

"@fortelabs google chat was painful for me too. just went with slack integration instead worked first try. for the token limit openrouter has way higher caps than direct anthropic api"
X Link 2026-02-12T00:00Z [---] 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

"@burkov the best part is "commit and push" as a single action with a mystery commit message. at least give me a preview before you tell my team i "refactored auth layer for improved modularity" when i actually just renamed a variable"
X Link 2026-02-12T01:46Z [---] 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 so most openclaw iterations are just scaffolding running tests checking output. route those through sonnet or haiku on openrouter way cheaper per token. only the final pass where it needs real architectural decisions goes to opus or gpt-5.2. went from $40/session to $8-12 with this split https://twitter.com/i/web/status/2021782248723558460 https://twitter.com/i/web/status/2021782248723558460"
X Link 2026-02-12T03:03Z [---] followers, [--] engagements

"@matt_slotnick wondering about this too. the irony would be brutal. AI was supposed to be the great equalizer and instead the companies with existing distribution just ship 3x faster"
X Link 2026-02-12T03:18Z [---] followers, [--] engagements

"@Hesamation hired dozens of people over the years. the frustrating truth is all of those things DO beat pure skills in the short run. but the people who stick around and grow skills compound. everything else has diminishing returns"
X Link 2026-02-12T03:30Z [---] followers, [---] engagements

"@emollick afaik mcp tools are the closest to a universal interface between code and cowork. skills are runtime-specific so they dont transfer. wrapping the imagegen call as an mcp server would let both environments call it. havent tested this specific flow in cowork tho so ymmv"
X Link 2026-02-12T05:53Z [---] followers, [---] engagements

"@Yuchenj_UW @steipete that growth curve is nuts but im always skeptical of stars as the metric. pytorch had its hockey stick moment too and took another [--] years before it was actually the default in production"
X Link 2026-02-12T06:02Z [---] followers, [----] engagements

"@thomasg_eth @sergeykarayev @banteg same arc. the mental model shifted from "writing code" to "reviewing and directing code" and honestly thats what senior devs were already doing with juniors. the junior just got mass 10x faster"
X Link 2026-02-12T06:07Z [---] followers, [---] engagements

"@davidfowl mIRC scripts were the original vibe coding tbh. custom color themes and /nick before css was even a thing"
X Link 2026-02-12T06:20Z [---] followers, [---] engagements

"glm-5. MIT licensed. zero nvidia gpus. [---] billion parameters. trained entirely on huawei chips. matches claude on coding benchmarks. apparently the chip export ban is working great lol"
X Link 2026-02-12T07:04Z [---] followers, [---] engagements

"@jarredsumner the most honest pr review ive ever seen. no nitpicks no "can we rename this variable" just. test"
X Link 2026-02-12T09:00Z [---] followers, [---] engagements

"@harjotsgill zero switching costs if youre just prompting. once you build actual workflows around one tho (configs mcp servers hooks) thats months of work you cant port. the agent is commodity the system around it isnt"
X Link 2026-02-12T10:08Z [---] 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

"@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

"@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

"@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

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

"@heygeorgekal mdfind. its macOS spotlight from terminal. faster than grep for finding stuff across projects and nobody seems to use it"
X Link 2026-02-12T16:33Z [---] 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

"@simonw @openclaw every openclaw tutorial is "let it run overnight and wake up to [--] completed repos" with zero mention of what happens to the repos on the receiving end of those PRs"
X Link 2026-02-12T18:03Z [---] 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

"@md_kadiwal [--] years running products. the ratio depends entirely on what you built. if its a screaming need the product IS the distribution channel because users tell each other. if its a nice-to-have then yeah most of your time goes to convincing people they need it"
X Link 2026-02-12T18:14Z [---] followers, [--] engagements

"@mattpocockuk the opus [---] observation matches what im seeing. less cleanup after each session. but "no entropy at all" assumes the model understands intent perfectly. right now the gap between what i meant and what the agent built is where the new entropy lives"
X Link 2026-02-12T19:38Z [---] followers, [---] engagements

"@simonw holiday dip probably inflated the percentage. but they also shipped agent teams vs code extension and opus [---] all in that [--] week window. lot of real product not just seasonal bounce"
X Link 2026-02-12T20:54Z [---] followers, [---] engagements

"that link panel on the right is basically apple pay for the web. one tap vs typing [--] digits is night and day for conversion. we switched to stripe last year and the saved card experience alone cut our checkout drop-off noticeably. paddle handles tax beautifully but the friction gap at payment time adds up fast https://twitter.com/i/web/status/2022052047961698422 https://twitter.com/i/web/status/2022052047961698422"
X Link 2026-02-12T20:55Z [---] 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

"@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

"@jessethanley been running products for [--] years. never once needed kafka. postgres queue + a cron job handles everything we throw at it. most teams skip the boring answer because it doesnt look impressive in architecture docs"
X Link 2026-02-12T23:18Z [---] followers, [---] engagements

"@TheAhmadOsman swe-bench numbers holding up against opus is impressive. if the API pricing is right this could be a real option for batch coding tasks where you dont need the full claude code harness around it"
X Link 2026-02-12T23:20Z [---] 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

"@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

"@fortelabs around $250. claude max does 90% of it openrouter for the rest. replaced like [--] saas subscriptions so net positive"
X Link 2026-02-13T00:47Z [---] 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

"@joshuajohnsonAI your dad was right lol. been rotating between different AI coding tools depending on the task for months. whats your split between the two"
X Link 2026-02-13T00:56Z [---] followers, [--] engagements

"@openclaw an IRC channel in [----]. beautiful"
X Link 2026-02-13T02:28Z [---] followers, [----] engagements

"@melvynxdev same. everyone around me is cursor or claude code. codex is more of a twitter thing from what ive seen"
X Link 2026-02-13T02:30Z [---] followers, [---] engagements

"@Anirudhgargi @lucas_montano fair for the happy path yea. but the second something breaks at 2am and you cant read the stack trace because theres no source code. thats when it matters. shipping fast isnt the same as shipping sustainably"
X Link 2026-02-13T02:33Z [---] followers, [--] engagements

"@joshuajohnsonAI wait UE5 + cursor how does it handle the unreal C++ stuff always figured game engine code would be too specialized for AI tools"
X Link 2026-02-13T02:34Z [---] followers, [--] engagements

"@batuhan the compaction loop kills momentum. you get it oriented it compacts then rediscovers the same files all over again. smaller context window would explain why spark does it more. ive been anchoring key context in project docs so it skips the rediscovery phase after compaction"
X Link 2026-02-13T02:37Z [---] 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

"@prajaybasu @antirez vs code extension handles files yeah. its the mcp integrations + hooks (pre/post action bash scripts) that dont have an equivalent yet. different workflows"
X Link 2026-02-13T03:50Z [---] followers, [--] engagements

"@Anirudhgargi @lucas_montano thats the bet yeah. question is whether you want your entire system dependent on one vendors AI always being available to fix itself. code at least gives you options"
X Link 2026-02-13T03:51Z [---] followers, [--] engagements

"@blader i give opus the "figure out whats broken" jobs and codex the "now go fix all [--] files" jobs. different beasts entirely lol"
X Link 2026-02-13T03:55Z [---] followers, [---] engagements

"@manuelmaly @antirez some of that works. but the mcp integrations talk to live services (supabase chrome external apis). mocking all of that into a sandbox is its own project lol"
X Link 2026-02-13T07:11Z [---] 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

"@yshaikh2001 convex if you need real-time. supabase if you want raw postgres power. been running both and they solve different problems honestly"
X Link 2026-02-13T07:31Z [---] followers, [---] engagements

"@martinicancer @prajaybasu @antirez fair point on mcp stand corrected. the hooks (bash scripts that run before/after every action) are the part without a codex equivalent yet. thats whats keeping me on claude code rn"
X Link 2026-02-13T09:49Z [---] 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

"@TheAhmadOsman weights dropping today gonna test it this weekend. whats the context window like thats usually where open source falls apart for real agent workflows. if it handles 100k+ well with that speed could actually replace some of my api calls"
X Link 2026-02-13T10:01Z [---] followers, [---] engagements

"depends on the project. for client work i still start with a quick wireframe because clients need something visual to approve. but for my own products straight to code every time now. with ai coding tools the feedback loop is faster than any mockup. describe what i want working version in minutes iterate from there. used to spend days in figma before writing a single line. now the prototype IS the code"
X Link 2026-02-13T10:10Z [---] 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

"@robinebers both are shipping tho. anthropic dropped claude code mcp protocol and the agent sdk back to back. they just dont have the same pr machine. openai is louder about it not necessarily faster"
X Link 2026-02-13T14:10Z [---] followers, [---] engagements

"@ivanfioravanti barely finished benchmarking k2.5 and m2.5 is already here. the pace of open weights right now is nuts. local inference getting competitive fast"
X Link 2026-02-13T14:20Z [---] followers, [--] engagements

"@robinebers agent teams and opus [---] dropped [--] days ago tho both pure dev features. feels like theyre expanding audience (cowork for PMs claude code for devs) rather than slowing down on either"
X Link 2026-02-13T15:17Z [---] followers, [--] engagements

"@gdb @ryancarson been switching between codex and claude code daily. UI is genuinely better for diff reviews thats not close. two things id love: less aggressive timeouts (losing context mid-task is painful) and some kind of extensibility layer for custom workflows"
X Link 2026-02-13T15:23Z [---] followers, [---] engagements

"@IndianTechGuide took trains across germany for years during university. fares were closer to france on that list service was. inconsistent lol. the part people miss here is india manufacturing these locally at icf chennai. thats what drives the cost gap not just cheaper labor"
X Link 2026-02-13T15:29Z [---] followers, [---] engagements

"@notgavn youre not doing anything wrong. opus just has better context awareness for existing codebases imo. i bounce between opus for anything structural and sonnet/codex for quick scaffolding. ends up being a workflow thing more than a loyalty thing"
X Link 2026-02-13T15:30Z [---] followers, [--] engagements

"@stats_feed the gap isnt about technology. nokia had smartphone prototypes years before iphone. its about willingness to cannibalize your own cash cow. most companies cant do it most founders cant either"
X Link 2026-02-13T15:35Z [---] followers, [---] engagements

"@giorgiottolina @antirez local fs yeah but hooks and skills arent there yet. thats whats keeping me on claude code rn. wont be surprised if codex catches up tho"
X Link 2026-02-13T16:43Z [---] 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

"@cristianuibar bolt is a good middle ground honestly. figma feels like overkill for most solo projects now"
X Link 2026-02-13T18:15Z [---] followers, [--] engagements

"@giorgiottolina @antirez the bash hook thing is more granular runs before/after every tool call not just at task level. but honestly codex is closing the gap fast so who knows in a month"
X Link 2026-02-13T18:17Z [---] followers, [--] engagements

"@cristianuibar same split honestly. gemini with search grounding is better for research and finding recent stuff. claude i lean on more when im actually building or need to reason through tricky architecture. different tools for different phases"
X Link 2026-02-13T18:29Z [---] followers, [--] engagements

"the 50x cheaper part is already happening. rebuilt [--] internal tools last month with claude code that wouldve been [--] weeks of contractor work. but the part nobody talks about is the review loop. agent gets 80% right then you burn [--] hours debugging the other 20% because you trusted the first output. reading code you didnt write is the actual bottleneck now not writing it https://twitter.com/i/web/status/2022413092169396713 https://twitter.com/i/web/status/2022413092169396713"
X Link 2026-02-13T20:50Z [---] followers, [---] engagements

"@PhilKiel ran a client campaign last month where reels got 4x the reach of static feed posts. same creative just reformatted. the platform is practically begging brands to go video"
X Link 2026-02-13T20:58Z [---] followers, [---] engagements

"@fortelabs been structuring my claude code workspace with projects/ knowledge/ resources/ without even naming it. basically reinvented PARA by accident lol"
X Link 2026-02-13T22:09Z [---] 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

"@divlams gpt-4o crashing 77% of the time while trying to flirt is extremely relatable"
X Link 2026-02-13T22:21Z [---] followers, [---] engagements

"@audibledata @antirez fair for most workflows yeah. my setup runs [--] mcp servers + browser automation locally tho doesnt sandbox cleanly"
X Link 2026-02-13T23:21Z [---] followers, [--] engagements

"@nummanali does each mirror get its own mcp server instances or do they share from the host"
X Link 2026-02-13T23:29Z [---] followers, [---] engagements

"@theo benchmarks dont test this and it shows"
X Link 2026-02-13T23:38Z [---] followers, [----] engagements

"@cyburns88 @bcherny claude code mostly. ssh into my server from the phone describe whats wrong come back to a fix. works surprisingly well for bugs where you know the expected behavior"
X Link 2026-02-14T00:56Z [---] followers, [---] engagements

"@RhysSullivan already happening. half my daily tools are mcp servers now no dashboard. supabase firecrawl stripe. the ui is just the agent"
X Link 2026-02-14T01:01Z [---] followers, [---] engagements

"@theo the ai reviewing its own UI in the screenshot is hilarious. if this actually pulls from custom design guidelines it would fix half the vibe coding complaints. any plans for custom skill packs"
X Link 2026-02-14T01:18Z [---] followers, [---] engagements

"@fortelabs the claude code route is genuinely the fastest. open terminal install claude code then literally just ask it to set up openclaw for you. it handles all the dependencies and config. took me maybe [--] min total. vanilla install works surprisingly well out of the box"
X Link 2026-02-14T02:38Z [---] followers, [---] engagements

"@gdb the "never published" bit is the whole story here. every other benchmark leaks into training data eventually. 6/10 on problems where pattern matching from the corpus isnt possible is a completely different kind of test"
X Link 2026-02-14T04:18Z [---] followers, [---] engagements

"@emollick the metacalculus estimate still says [----] lol. feels like the benchmarks are tracking a version of AI that stopped being the frontier about [--] years ago. by the time we formally check every box the conversation will have moved to completely different capabilities"
X Link 2026-02-14T04:19Z [---] followers, [---] engagements

"the piano analogy undersells it tbh. piano stays the same instrument forever. AI rewrites itself every few months so youre learning and unlearning simultaneously. the real skill isnt prompting mechanics its problem decomposition and knowing what to build. that part compounds regardless of model improvements https://twitter.com/i/web/status/2022527348546076968 https://twitter.com/i/web/status/2022527348546076968"
X Link 2026-02-14T04:24Z [---] followers, [---] engagements

"@allgarbled @growing_daniel lol claude just casually adding labubu to the tech stack"
X Link 2026-02-14T04:36Z [---] followers, [---] engagements

"the solo founder fantasy misses that building was never the bottleneck. distribution trust and domain expertise were. ive been bootstrapping for [--] years and the hardest parts of running profitable businesses had nothing to do with code. AI agents make building cheaper but they dont solve the "how do i get [----] people to care" problem. if anything when everyone can build anything the stuff that cant be automated becomes 10x more valuable"
X Link 2026-02-14T04:38Z [---] followers, [----] engagements

"@tenobrus the stuck loop happens because models get locked into one reasoning path and just keep trying variations of the same broken assumption. different model = different starting priors. been swapping between opus and codex specifically for these moments"
X Link 2026-02-14T06:21Z [---] followers, [--] engagements

"@tysonhutchins_ @RobertJBye same. i run sessions on two macs and keeping context in sync is a mess. ssh + tmux kinda works but you lose all the tooling. what id want is something like vs code remote but for claude code model context stays on one machine and you just attach from wherever"
X Link 2026-02-14T06:32Z [---] followers, [---] engagements

"@DjMolehill @theo at that point youre basically whispering to your laptop"
X Link 2026-02-14T09:12Z [---] followers, [---] engagements

"@robinebers wait they work on claude code for web too been installing everything globally like a fool this whole time"
X Link 2026-02-14T09:18Z [---] followers, [--] engagements

"run an agency. the code was never the expensive part. scoping what to build managing stakeholder disagreements iterating on feedback loops and keeping things running after launch is 80% of the work. ai handles the typing faster but someone still needs to figure out what to type"
X Link 2026-02-14T11:38Z [---] followers, [--] engagements

"@Glenkurt_ @tenobrus honestly cold outreach + showing up where people already ask questions. goausbildung first users came from answering german education questions on reddit and quora that nobody else bothered with. zero ads. just showed up consistently where demand already existed"
X Link 2026-02-14T13:57Z [---] followers, [--] engagements

"@kr0der next.js + supabase + tailwind + hetzner with coolify. swapped vercel for self hosted deploys last year same dx once you set it up. convex when something needs realtime. total hosting cost across [--] projects is like $20/mo"
X Link 2026-02-14T18:48Z [---] followers, [--] engagements

"@vincent_vancode this is basically what MCP is trying to do but from the tool side. an file at the root of every website would be huge tho. like robots.txt but for agents instead of crawlers. someone should just ship a spec and see if it sticks http://index.ai http://index.ai"
X Link 2026-02-15T02:57Z [---] followers, [--] engagements

"@LeeLeepenkman this is basically how i work already. fast model for scaffolding opus when the task needs reasoning or architecture. the routing is simpler than people think just match task complexity to model weight. cuts token costs by like 40% across projects"
X Link 2026-02-15T04:37Z [---] followers, [--] engagements

"@martinmuzatko cloudflare R2. s3-compatible API so you just swap the endpoint egress is completely free and pricing is predictable. been running it across [--] projects for about a year zero surprises on the bill. way better than minio and you skip the aws cost anxiety entirely"
X Link 2026-02-15T07:17Z [---] followers, [--] engagements

"@eliana_jordan most people treat X payouts like a bonus check. reinvesting into tools that improve your next piece of content is how you actually turn this into a flywheel and not just a side hustle"
X Link 2026-02-15T07:21Z [---] followers, [--] engagements

"@lochan_twt claude code subscription before a small investment tells you everything about [----] priorities and honestly same"
X Link 2026-02-15T07:22Z [---] followers, [----] engagements

"@kr0der fair point actually. for me it was more about running [--] sites on one box. $7/mo total vs $20/mo per project adds up fast when you keep shipping"
X Link 2026-02-15T09:15Z [---] followers, [--] engagements

"@swyx @joshuaday seen this exact pattern with every HN thread. [--] comments about your database choice zero about the problem you solve. engineers will argue about ORMs for hours but never ask whos paying and why"
X Link 2026-02-15T09:23Z [---] followers, [--] engagements

"@kapilansh_twt we run a youtube channel for our edtech product. [---] view videos bring more signups than 50k view ones because the intent is completely different"
X Link 2026-02-15T12:29Z [---] followers, [--] engagements

"@1Umairshaikh stripe dashboard at 2am is a whole different sport"
X Link 2026-02-15T14:23Z [---] followers, [--] 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

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