#  @nick_tikhonov @nt on Farcaster @nt on Farcaster posts on X about llm, farcaster, gpt2, protocol the most. They currently have [-----] followers and [--] posts still getting attention that total [--] engagements in the last [--] hours. ### Engagements: [--] [#](/creator/twitter::3040961572/interactions)  - [--] Week [---] +57% - [--] Months [------] +9,294% - [--] Year [------] +10,223% ### Mentions: [--] [#](/creator/twitter::3040961572/posts_active)  ### Followers: [-----] [#](/creator/twitter::3040961572/followers)  - [--] Week [-----] +0.48% - [--] Months [-----] +2.50% - [--] Year [-----] +12% ### CreatorRank: undefined [#](/creator/twitter::3040961572/influencer_rank)  ### Social Influence **Social category influence** [social networks](/list/social-networks) [stocks](/list/stocks) [technology brands](/list/technology-brands) [finance](/list/finance) [countries](/list/countries) **Social topic influence** [llm](/topic/llm), [farcaster](/topic/farcaster), [gpt2](/topic/gpt2), [protocol](/topic/protocol), [topics](/topic/topics), [acquisition](/topic/acquisition), [age of](/topic/age-of), [prisma](/topic/prisma), [prod](/topic/prod), [ai](/topic/ai) **Top accounts mentioned or mentioned by** [@seansnd](/creator/undefined) [@tootallnate](/creator/undefined) [@pranaygp](/creator/undefined) [@twilio](/creator/undefined) [@deepgramai](/creator/undefined) [@elevenlabsio](/creator/undefined) [@brianarmstrong](/creator/undefined) [@thenetworkstate](/creator/undefined) [@obsdmd](/creator/undefined) [@pandelisz](/creator/undefined) [@andymmmmc](/creator/undefined) [@quantstamp](/creator/undefined) [@littleatlasxyz](/creator/undefined) [@originlondon](/creator/undefined) [@8270eth](/creator/undefined) [@punk6529](/creator/undefined) [@imgyf](/creator/undefined) [@jamesoleary32](/creator/undefined) [@culturecrit](/creator/undefined) [@pierredewulf](/creator/undefined) **Top assets mentioned** [GPT2 (GPT2)](/topic/gpt2) ### Top Social Posts Top posts by engagements in the last [--] hours "Get you an angel investor/advisor that reigns you in with custom memes 😂💯- @andymmmmc" [X Link](https://x.com/nick_tikhonov/status/1555318247876067328) 2022-08-04T22:22Z [----] followers, [--] engagements "Whats cool is since the Farcaster platform is actually a protocol its easy to scrape and crunch the numbers on the community: 📈 7k+ users 🔧 2.5k founders/builders 💰 [---] investors 🗓 3.5k posted in the last [--] days" [X Link](https://x.com/nick_tikhonov/status/1611757085510737921) 2023-01-07T16:10Z [----] followers, [---] engagements "@Quantstamp This looks awesome 🙌" [X Link](https://x.com/nick_tikhonov/status/1668183095353155589) 2023-06-12T09:06Z [----] followers, [--] engagements "@brian_armstrong We've been working on @thenetworkstate operating dashboard for the past year - starting by helping tokenised communities coordinate IRL to help them move along the axis from online to physical. In some cases we've increased local coordination for these communities 10x" [X Link](https://x.com/NFTikhonov/status/1698974522982998328) 2023-09-05T08:21Z [----] followers, [--] engagements "@brian_armstrong @thenetworkstate Here's a deeper look into what we're working on @littleatlas_xyz" [X Link](https://x.com/NFTikhonov/status/1698982557272908228) 2023-09-05T08:53Z [----] followers, [--] engagements "Overheard @OriginLondon - I bought this customer Nandos one time and I got so much quality feedback from there on out" [X Link](https://x.com/nick_tikhonov/status/1699087438885355668) 2023-09-05T15:49Z [----] followers, [---] engagements "I found that this additional cost made it hard to switch from Notes to @obsdmd initially: The added cost of organising knowledge is immediate but you don't feel the compounding value of a having an interconnected knowledge base until you've built a graph of a certain size" [X Link](https://x.com/nick_tikhonov/status/1735650137161019824) 2023-12-15T13:16Z [----] followers, [--] engagements "After [--] days of using @obsdmd I've collected [---] notes across [--] topics. I am now starting to feel the value by: - jumping to random notes and navigating the graph to refresh my memory - rewriting/simplifying frequently accessed notes to revise my thinking" [X Link](https://x.com/NFTikhonov/status/1735650138847166747) 2023-12-15T13:16Z [----] followers, [--] engagements "Two months into using a Zettelkasten / connected notes / evergreen notes every day ✍ ⚡ I'm starting to notice a big difference in the quality of my thinking and arguments especially when discussing what I've learned with others ✍ the rather new and alien note-taking habit has mostly solidified and reached a pleasant equilibrium - the first month was full of confusion and wilful experimentation - but by now I'm writing notes pretty much on autopilot 📈 with [---] concept notes 35k words and 1k links I spend an increasing amount of time reading my own writing - exploring reviewing and finding new" [X Link](https://x.com/nick_tikhonov/status/1746885988943605997) 2024-01-15T13:24Z [----] followers, [---] engagements "I'm working on a note-taking system that helps founders spend more time thinking. It uses AI to automatically find relevant connections for each new note you enter eliminating the need to maintain complex Roam/Obsidian-style graphs. Looking for beta testers" [X Link](https://x.com/nick_tikhonov/status/1753448405483782589) 2024-02-02T16:00Z [----] followers, [---] engagements "@8270eth I've built one with Obsidian for [--] months grinding through the learning curve. It's tough but valuable I'm building an AI version that greatly reduces the learning curve whilst keeping the value. I'm working on a note-taking system that helps founders spend more time thinking. It uses AI to automatically find relevant connections for each new note you enter eliminating the need to maintain complex Roam/Obsidian-style graphs. Looking for beta testers https://t.co/rRg7Dqrlvi I'm working on a note-taking system that helps founders spend more time thinking. It uses AI to automatically" [X Link](https://x.com/nick_tikhonov/status/1753754968576688419) 2024-02-03T12:19Z [----] followers, [--] engagements "@punk6529 you should join farcaster" [X Link](https://x.com/nick_tikhonov/status/1789517623182405752) 2024-05-12T04:47Z [----] followers, [--] engagements "Scott I love your content and want to support you and read your thesis. But your book purchase flow is at best tastelessly annoying and at worst predatory and manipulative. I bought your book and ebook for $20 then AFTER payment I had to scroll through an entire page of upselling (annoying but acceptable) and a 10-minute video that I could not skip sleezily upselling me on your course like and telling me that I would FAIL if I do not take it. The video cannot be skipped and I am told that my order could fail if I close the page. You are trying to grow a community of intelligent independent" [X Link](https://x.com/nick_tikhonov/status/1825846025623314884) 2024-08-20T10:43Z [----] followers, [--] engagements "two sides of Silicon Valley: first: "let's hire a dev to ship this" do things that dont scale enterprise SaaS user interviews product/business CEO. second: fuck around and find out hackathons tinkering research and deep tech engineer as CEO" [X Link](https://x.com/nick_tikhonov/status/1833123752617422851) 2024-09-09T12:42Z [----] followers, [---] engagements "all news is fake. how can you expect reality from a system that optimises for engagement reality is boring. news constructs the kind of reality we cant afford to take our eyes off of" [X Link](https://x.com/nick_tikhonov/status/1840372509318820292) 2024-09-29T12:46Z [----] followers, [---] engagements "@jamesoleary32 Yo thats 🔥 congrats" [X Link](https://x.com/nick_tikhonov/status/1844666465829462223) 2024-10-11T09:08Z [----] followers, [--] engagements "@Culture_Crit Plato not Socrates 🤦" [X Link](https://x.com/nick_tikhonov/status/1853022719354298797) 2024-11-03T10:33Z [----] followers, [--] engagements "@PierreDeWulf i think you forgot to count sales tax on the spent money" [X Link](https://x.com/nick_tikhonov/status/1854595134333088137) 2024-11-07T18:41Z [----] followers, [----] engagements "day [--] of learning to train LLMs: I have a 100m parameter (GPT-2 small eqiv.) model training on roughly 1.2GB of public domain books generating what sounds like a young child trying to use complicated words 🤣" [X Link](https://x.com/nick_tikhonov/status/1857420286196355122) 2024-11-15T13:48Z [----] followers, [---] engagements "day [--] of learning to train LLMs. The 11th iteration is a replica of GPT-2 small with 168m parameters. starting by training it on a few GB of public domain books but will eventually learn how to scale this up by 100x-1000x. here are the first few thousand training iterations:" [X Link](https://x.com/nick_tikhonov/status/1858822755287806373) 2024-11-19T10:40Z [----] followers, [--] engagements "@RKTbyClanker @clankfun @clankeronbase @farcaster_xyz mistake on our end - lets get you added 💯" [X Link](https://x.com/nick_tikhonov/status/1866944963352592590) 2024-12-11T20:35Z [----] followers, [--] engagements "@elonmuskTN [--]. [--] buds = [--] [--] bud = [--] [--]. [--] guy is [--] (5 + [--] + [--] = 20) [--]. [--] watches is [--] (4 + [--] + [--] = 13) [--]. [--] bud + [--] guy [--] buds watch * [--] watch = [--] + ((10 + [--] + 2) * 2) = 39" [X Link](https://x.com/nick_tikhonov/status/1873049155045867976) 2024-12-28T16:51Z [----] followers, [--] engagements "It blew my mind that there is no integrated way to vibe code SQL queries. so I vibe coded SnapQL: a desktop app that lets you query any postgres db with AI in seconds ⚡" [X Link](https://x.com/nick_tikhonov/status/1935409780522352773) 2025-06-18T18:50Z [----] followers, [---] engagements "comment if you'd like to beta test it today 👇" [X Link](https://x.com/nick_tikhonov/status/1935409858808799518) 2025-06-18T18:50Z [----] followers, [--] engagements "@KintuLabs here's the GitHub repo https://github.com/NickTikhonov/snap-ql https://github.com/NickTikhonov/snap-ql" [X Link](https://x.com/nick_tikhonov/status/1935625183588475170) 2025-06-19T09:06Z [----] followers, [--] engagements "https://github.com/NickTikhonov/snap-ql https://github.com/NickTikhonov/snap-ql" [X Link](https://x.com/nick_tikhonov/status/1935647414968561681) 2025-06-19T10:34Z [----] followers, [--] engagements "@Beareka It's pretty good" [X Link](https://x.com/nick_tikhonov/status/1935647568316223863) 2025-06-19T10:35Z [----] followers, [--] engagements "@serhii_be postgres only for now but feel free to raise an issue/PR: https://github.com/NickTikhonov/snap-ql https://github.com/NickTikhonov/snap-ql" [X Link](https://x.com/nick_tikhonov/status/1935647626457657706) 2025-06-19T10:35Z [----] followers, [--] engagements "you can now vibe code db queries 👀" [X Link](https://x.com/nick_tikhonov/status/1935648563226157193) 2025-06-19T10:39Z [----] followers, [---] engagements "fully local and open source: https://github.com/NickTikhonov/snap-ql https://github.com/NickTikhonov/snap-ql" [X Link](https://x.com/nick_tikhonov/status/1935648661666660396) 2025-06-19T10:39Z [----] followers, [--] engagements "when thinking about building products I am now super suspicious if I have an idea that takes more than a few days to ship. anything more ambitious than that probably has a set of "stepping-stone" features that act as intermediate for validation and user acquisition" [X Link](https://x.com/nick_tikhonov/status/1936144353686131117) 2025-06-20T19:29Z [----] followers, [---] engagements "why small teams win in the age of AI https://ntik.me/posts/small-teams https://ntik.me/posts/small-teams" [X Link](https://x.com/nick_tikhonov/status/1955361016940040337) 2025-08-12T20:09Z [----] followers, [--] engagements "@haydenbleasel @strehldev Hey hayden next-forge doesnt work at all at the moment. Literally clean deploy on Vercel breaks because of prisma openssl compat" [X Link](https://x.com/nick_tikhonov/status/1984166698048745730) 2025-10-31T07:52Z [----] followers, [--] engagements "@pranaygp @sean_snd @TooTallNate got it working great in prod only issue is sleep() seems unreliable - e.g. if I sleep for [--] minutes my workflows actually sleep quite a bit longer" [X Link](https://x.com/nick_tikhonov/status/1985803687751561565) 2025-11-04T20:17Z [----] followers, [---] engagements "@TooTallNate @pranaygp @sean_snd great will try that. thank you both" [X Link](https://x.com/nick_tikhonov/status/1985815334809378919) 2025-11-04T21:04Z [----] followers, [---] engagements "fully migrated to workflows - our use case are AI agents that execute over a multiple-day time frame making multiple outbound voice calls and processing the results before: - scheduling service - queues - workers - cron jobs now: - [--] functions in one file 🤯 @nick_tikhonov @sean_snd @TooTallNate Go for it Were already using at Vercel in production for 1M workflows a day :) Well keep adding more to the APIs as there are a lot of features we want to ship - but the essentials are stable @nick_tikhonov @sean_snd @TooTallNate Go for it Were already using at Vercel in production for 1M workflows a" [X Link](https://x.com/nick_tikhonov/status/1985971284577050699) 2025-11-05T07:23Z [----] followers, 78.7K engagements "@TooTallNate @pranaygp @sean_snd one of the things I also had to do was rewrite webhook urls to use the production domain as the vercel domains seemed to have an extra auth layer similar to how preview deployments do could this be something I misconfigured cc @pranaygp @TooTallNate" [X Link](https://x.com/nick_tikhonov/status/1986075673874624795) 2025-11-05T14:18Z [----] followers, [--] engagements "those preferences are what define a product. any product can be expressed as the sum of all feedback it has received from its users. if you were to collate that feedback and pipe it into the LLM of [--] years from now youd get that product very different from farming/spinning wool/etc where the act of production has nothing to do with consumption" [X Link](https://x.com/nick_tikhonov/status/1986718108921254378) 2025-11-07T08:51Z [----] followers, [--] engagements "genuine wow moment with Gemini 3: I jammed for [--] hours with Gemini and built a fully functional Vapi/Elevenlabs phone agent clone that orchestrates TTS STT LLM and turn-taking. you can just build your own stack now" [X Link](https://x.com/nick_tikhonov/status/1991437857333080110) 2025-11-20T09:25Z [----] followers, [---] engagements "this is actually a very interesting problem to solve: you have a socket open with Twilio which lets you send and receive voice frames every couple of ms. then you have to decide whether the user is speaking or not you do this by piping the audio frames to a transcription service and also a statistical model that measures the instantaneous volume of the audio signal. this catches the exact point the user starts and stops speaking. very important when the user stops speaking we feed the transcript to an LLM and start streaming it's output to a TTS service (like 11labs). we get back audio frames" [X Link](https://x.com/nick_tikhonov/status/1991442037153890317) 2025-11-20T09:42Z [----] followers, [--] engagements "I think it's a post-rationalisation of emotions I bet europeans/americans find every reason to love Japan because of it being explicitly harmless to them post ww2 I bet europeans will find every reason to dislike China because of how much it has progressed economically in the last decade" [X Link](https://x.com/nick_tikhonov/status/1996583009038864703) 2025-12-04T14:10Z [----] followers, 20.8K engagements "@levelsio I think this thought lives in most people's shadow i.e. extremely uncomfortable for most people to consciously confront it - esp with nationalist narratives/conditioning" [X Link](https://x.com/nick_tikhonov/status/1996583863078781300) 2025-12-04T14:14Z [----] followers, [----] engagements "@micahstubbs interrupt handling is the crux - I was playing around with a few statistical models gemini did a decent job but I eventually took over to iterate on this part by hand" [X Link](https://x.com/nick_tikhonov/status/1996622082994844062) 2025-12-04T16:46Z [----] followers, [--] engagements "@PandelisZ @twilio @DeepgramAI @elevenlabsio there are a lot of factors but LLM latency accounts for about half of the total. really depends what you're going to do with the model - 4o-mini is fine for this project but have found it hard to steer towards reliable performance in production" [X Link](https://x.com/nick_tikhonov/status/2019695497699856674) 2026-02-06T08:51Z [----] followers, [--] engagements "the trick to making voice agents fast is pipelining everything first pipeline: audio packets - speech to text - turn-taking model second: LLM - text to speech - encoding - output here's a render of my current latency. bear in mind I'm running this locally from a wooden hut in the mountains in Turkey - it should be significantly faster once I host it https://twitter.com/i/web/status/2019801839240159411 https://twitter.com/i/web/status/2019801839240159411" [X Link](https://x.com/nick_tikhonov/status/2019801839240159411) 2026-02-06T15:54Z [----] followers, [---] engagements "@seltzergenius @twilio @DeepgramAI @elevenlabsio gotchu this motivated me to post the next update" [X Link](https://x.com/nick_tikhonov/status/2019802006425145755) 2026-02-06T15:54Z [----] followers, [--] engagements "this architecture gets you to a nearly SOTA voice agent (a la Vapi/ElevenLabs) but with much more control over the orchestration. @DeepgramAI's Flux handles STT and turn taking + a pipeline for LLM and TTS managed based on Flux events" [X Link](https://x.com/nick_tikhonov/status/2020035798473556126) 2026-02-07T07:23Z [----] followers, [---] engagements "day [--] learning to build voice agent infra from scratch: put together a VAD @twilio @DeepgramAI (SST) o4-mini @elevenlabsio (TTS) into an event-based loop that coordinates listening and speaking main issue right now is latency and quality of turn-taking/interruptions. not quite as good as Vapi/Elevenlabs off-the-shelf next I'm going to try using Deepgram's Flux for both STT and endpointing https://twitter.com/i/web/status/2019693047022289387 https://twitter.com/i/web/status/2019693047022289387" [X Link](https://x.com/nick_tikhonov/status/2019693047022289387) 2026-02-06T08:42Z [----] followers, [---] engagements "just beat @Vapi_AI E2E latency with my own voice agent orchestrator. had to host everything just right to get the numbers this low Vapi's reported: 840ms mine: 690ms + 100ms (twilio) = 790ms that 50ms is a big deal in voice write-up incoming" [X Link](https://x.com/nick_tikhonov/status/2020790064272355679) 2026-02-09T09:21Z [----] followers, [---] engagements Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing
@nick_tikhonov @nt on Farcaster
@nt on Farcaster posts on X about llm, farcaster, gpt2, protocol the most. They currently have [-----] followers and [--] posts still getting attention that total [--] engagements in the last [--] hours.
Engagements: [--] #
- [--] Week [---] +57%
- [--] Months [------] +9,294%
- [--] Year [------] +10,223%
Mentions: [--] #
Followers: [-----] #
- [--] Week [-----] +0.48%
- [--] Months [-----] +2.50%
- [--] Year [-----] +12%
CreatorRank: undefined #
Social Influence
Social category influence social networks stocks technology brands finance countries
Social topic influence llm, farcaster, gpt2, protocol, topics, acquisition, age of, prisma, prod, ai
Top accounts mentioned or mentioned by @seansnd @tootallnate @pranaygp @twilio @deepgramai @elevenlabsio @brianarmstrong @thenetworkstate @obsdmd @pandelisz @andymmmmc @quantstamp @littleatlasxyz @originlondon @8270eth @punk6529 @imgyf @jamesoleary32 @culturecrit @pierredewulf
Top assets mentioned GPT2 (GPT2)
Top Social Posts
Top posts by engagements in the last [--] hours
"Get you an angel investor/advisor that reigns you in with custom memes 😂💯- @andymmmmc"
X Link 2022-08-04T22:22Z [----] followers, [--] engagements
"Whats cool is since the Farcaster platform is actually a protocol its easy to scrape and crunch the numbers on the community: 📈 7k+ users 🔧 2.5k founders/builders 💰 [---] investors 🗓 3.5k posted in the last [--] days"
X Link 2023-01-07T16:10Z [----] followers, [---] engagements
"@Quantstamp This looks awesome 🙌"
X Link 2023-06-12T09:06Z [----] followers, [--] engagements
"@brian_armstrong We've been working on @thenetworkstate operating dashboard for the past year - starting by helping tokenised communities coordinate IRL to help them move along the axis from online to physical. In some cases we've increased local coordination for these communities 10x"
X Link 2023-09-05T08:21Z [----] followers, [--] engagements
"@brian_armstrong @thenetworkstate Here's a deeper look into what we're working on @littleatlas_xyz"
X Link 2023-09-05T08:53Z [----] followers, [--] engagements
"Overheard @OriginLondon - I bought this customer Nandos one time and I got so much quality feedback from there on out"
X Link 2023-09-05T15:49Z [----] followers, [---] engagements
"I found that this additional cost made it hard to switch from Notes to @obsdmd initially: The added cost of organising knowledge is immediate but you don't feel the compounding value of a having an interconnected knowledge base until you've built a graph of a certain size"
X Link 2023-12-15T13:16Z [----] followers, [--] engagements
"After [--] days of using @obsdmd I've collected [---] notes across [--] topics. I am now starting to feel the value by: - jumping to random notes and navigating the graph to refresh my memory - rewriting/simplifying frequently accessed notes to revise my thinking"
X Link 2023-12-15T13:16Z [----] followers, [--] engagements
"Two months into using a Zettelkasten / connected notes / evergreen notes every day ✍ ⚡ I'm starting to notice a big difference in the quality of my thinking and arguments especially when discussing what I've learned with others ✍ the rather new and alien note-taking habit has mostly solidified and reached a pleasant equilibrium - the first month was full of confusion and wilful experimentation - but by now I'm writing notes pretty much on autopilot 📈 with [---] concept notes 35k words and 1k links I spend an increasing amount of time reading my own writing - exploring reviewing and finding new"
X Link 2024-01-15T13:24Z [----] followers, [---] engagements
"I'm working on a note-taking system that helps founders spend more time thinking. It uses AI to automatically find relevant connections for each new note you enter eliminating the need to maintain complex Roam/Obsidian-style graphs. Looking for beta testers"
X Link 2024-02-02T16:00Z [----] followers, [---] engagements
"@8270eth I've built one with Obsidian for [--] months grinding through the learning curve. It's tough but valuable I'm building an AI version that greatly reduces the learning curve whilst keeping the value. I'm working on a note-taking system that helps founders spend more time thinking. It uses AI to automatically find relevant connections for each new note you enter eliminating the need to maintain complex Roam/Obsidian-style graphs. Looking for beta testers https://t.co/rRg7Dqrlvi I'm working on a note-taking system that helps founders spend more time thinking. It uses AI to automatically"
X Link 2024-02-03T12:19Z [----] followers, [--] engagements
"@punk6529 you should join farcaster"
X Link 2024-05-12T04:47Z [----] followers, [--] engagements
"Scott I love your content and want to support you and read your thesis. But your book purchase flow is at best tastelessly annoying and at worst predatory and manipulative. I bought your book and ebook for $20 then AFTER payment I had to scroll through an entire page of upselling (annoying but acceptable) and a 10-minute video that I could not skip sleezily upselling me on your course like and telling me that I would FAIL if I do not take it. The video cannot be skipped and I am told that my order could fail if I close the page. You are trying to grow a community of intelligent independent"
X Link 2024-08-20T10:43Z [----] followers, [--] engagements
"two sides of Silicon Valley: first: "let's hire a dev to ship this" do things that dont scale enterprise SaaS user interviews product/business CEO. second: fuck around and find out hackathons tinkering research and deep tech engineer as CEO"
X Link 2024-09-09T12:42Z [----] followers, [---] engagements
"all news is fake. how can you expect reality from a system that optimises for engagement reality is boring. news constructs the kind of reality we cant afford to take our eyes off of"
X Link 2024-09-29T12:46Z [----] followers, [---] engagements
"@jamesoleary32 Yo thats 🔥 congrats"
X Link 2024-10-11T09:08Z [----] followers, [--] engagements
"@Culture_Crit Plato not Socrates 🤦"
X Link 2024-11-03T10:33Z [----] followers, [--] engagements
"@PierreDeWulf i think you forgot to count sales tax on the spent money"
X Link 2024-11-07T18:41Z [----] followers, [----] engagements
"day [--] of learning to train LLMs: I have a 100m parameter (GPT-2 small eqiv.) model training on roughly 1.2GB of public domain books generating what sounds like a young child trying to use complicated words 🤣"
X Link 2024-11-15T13:48Z [----] followers, [---] engagements
"day [--] of learning to train LLMs. The 11th iteration is a replica of GPT-2 small with 168m parameters. starting by training it on a few GB of public domain books but will eventually learn how to scale this up by 100x-1000x. here are the first few thousand training iterations:"
X Link 2024-11-19T10:40Z [----] followers, [--] engagements
"@RKTbyClanker @clankfun @clankeronbase @farcaster_xyz mistake on our end - lets get you added 💯"
X Link 2024-12-11T20:35Z [----] followers, [--] engagements
"@elonmuskTN [--]. [--] buds = [--] [--] bud = [--] [--]. [--] guy is [--] (5 + [--] + [--] = 20) [--]. [--] watches is [--] (4 + [--] + [--] = 13) [--]. [--] bud + [--] guy [--] buds watch * [--] watch = [--] + ((10 + [--] + 2) * 2) = 39"
X Link 2024-12-28T16:51Z [----] followers, [--] engagements
"It blew my mind that there is no integrated way to vibe code SQL queries. so I vibe coded SnapQL: a desktop app that lets you query any postgres db with AI in seconds ⚡"
X Link 2025-06-18T18:50Z [----] followers, [---] engagements
"comment if you'd like to beta test it today 👇"
X Link 2025-06-18T18:50Z [----] followers, [--] engagements
"@KintuLabs here's the GitHub repo https://github.com/NickTikhonov/snap-ql https://github.com/NickTikhonov/snap-ql"
X Link 2025-06-19T09:06Z [----] followers, [--] engagements
"https://github.com/NickTikhonov/snap-ql https://github.com/NickTikhonov/snap-ql"
X Link 2025-06-19T10:34Z [----] followers, [--] engagements
"@Beareka It's pretty good"
X Link 2025-06-19T10:35Z [----] followers, [--] engagements
"@serhii_be postgres only for now but feel free to raise an issue/PR: https://github.com/NickTikhonov/snap-ql https://github.com/NickTikhonov/snap-ql"
X Link 2025-06-19T10:35Z [----] followers, [--] engagements
"you can now vibe code db queries 👀"
X Link 2025-06-19T10:39Z [----] followers, [---] engagements
"fully local and open source: https://github.com/NickTikhonov/snap-ql https://github.com/NickTikhonov/snap-ql"
X Link 2025-06-19T10:39Z [----] followers, [--] engagements
"when thinking about building products I am now super suspicious if I have an idea that takes more than a few days to ship. anything more ambitious than that probably has a set of "stepping-stone" features that act as intermediate for validation and user acquisition"
X Link 2025-06-20T19:29Z [----] followers, [---] engagements
"why small teams win in the age of AI https://ntik.me/posts/small-teams https://ntik.me/posts/small-teams"
X Link 2025-08-12T20:09Z [----] followers, [--] engagements
"@haydenbleasel @strehldev Hey hayden next-forge doesnt work at all at the moment. Literally clean deploy on Vercel breaks because of prisma openssl compat"
X Link 2025-10-31T07:52Z [----] followers, [--] engagements
"@pranaygp @sean_snd @TooTallNate got it working great in prod only issue is sleep() seems unreliable - e.g. if I sleep for [--] minutes my workflows actually sleep quite a bit longer"
X Link 2025-11-04T20:17Z [----] followers, [---] engagements
"@TooTallNate @pranaygp @sean_snd great will try that. thank you both"
X Link 2025-11-04T21:04Z [----] followers, [---] engagements
"fully migrated to workflows - our use case are AI agents that execute over a multiple-day time frame making multiple outbound voice calls and processing the results before: - scheduling service - queues - workers - cron jobs now: - [--] functions in one file 🤯 @nick_tikhonov @sean_snd @TooTallNate Go for it Were already using at Vercel in production for 1M workflows a day :) Well keep adding more to the APIs as there are a lot of features we want to ship - but the essentials are stable @nick_tikhonov @sean_snd @TooTallNate Go for it Were already using at Vercel in production for 1M workflows a"
X Link 2025-11-05T07:23Z [----] followers, 78.7K engagements
"@TooTallNate @pranaygp @sean_snd one of the things I also had to do was rewrite webhook urls to use the production domain as the vercel domains seemed to have an extra auth layer similar to how preview deployments do could this be something I misconfigured cc @pranaygp @TooTallNate"
X Link 2025-11-05T14:18Z [----] followers, [--] engagements
"those preferences are what define a product. any product can be expressed as the sum of all feedback it has received from its users. if you were to collate that feedback and pipe it into the LLM of [--] years from now youd get that product very different from farming/spinning wool/etc where the act of production has nothing to do with consumption"
X Link 2025-11-07T08:51Z [----] followers, [--] engagements
"genuine wow moment with Gemini 3: I jammed for [--] hours with Gemini and built a fully functional Vapi/Elevenlabs phone agent clone that orchestrates TTS STT LLM and turn-taking. you can just build your own stack now"
X Link 2025-11-20T09:25Z [----] followers, [---] engagements
"this is actually a very interesting problem to solve: you have a socket open with Twilio which lets you send and receive voice frames every couple of ms. then you have to decide whether the user is speaking or not you do this by piping the audio frames to a transcription service and also a statistical model that measures the instantaneous volume of the audio signal. this catches the exact point the user starts and stops speaking. very important when the user stops speaking we feed the transcript to an LLM and start streaming it's output to a TTS service (like 11labs). we get back audio frames"
X Link 2025-11-20T09:42Z [----] followers, [--] engagements
"I think it's a post-rationalisation of emotions I bet europeans/americans find every reason to love Japan because of it being explicitly harmless to them post ww2 I bet europeans will find every reason to dislike China because of how much it has progressed economically in the last decade"
X Link 2025-12-04T14:10Z [----] followers, 20.8K engagements
"@levelsio I think this thought lives in most people's shadow i.e. extremely uncomfortable for most people to consciously confront it - esp with nationalist narratives/conditioning"
X Link 2025-12-04T14:14Z [----] followers, [----] engagements
"@micahstubbs interrupt handling is the crux - I was playing around with a few statistical models gemini did a decent job but I eventually took over to iterate on this part by hand"
X Link 2025-12-04T16:46Z [----] followers, [--] engagements
"@PandelisZ @twilio @DeepgramAI @elevenlabsio there are a lot of factors but LLM latency accounts for about half of the total. really depends what you're going to do with the model - 4o-mini is fine for this project but have found it hard to steer towards reliable performance in production"
X Link 2026-02-06T08:51Z [----] followers, [--] engagements
"the trick to making voice agents fast is pipelining everything first pipeline: audio packets - speech to text - turn-taking model second: LLM - text to speech - encoding - output here's a render of my current latency. bear in mind I'm running this locally from a wooden hut in the mountains in Turkey - it should be significantly faster once I host it https://twitter.com/i/web/status/2019801839240159411 https://twitter.com/i/web/status/2019801839240159411"
X Link 2026-02-06T15:54Z [----] followers, [---] engagements
"@seltzergenius @twilio @DeepgramAI @elevenlabsio gotchu this motivated me to post the next update"
X Link 2026-02-06T15:54Z [----] followers, [--] engagements
"this architecture gets you to a nearly SOTA voice agent (a la Vapi/ElevenLabs) but with much more control over the orchestration. @DeepgramAI's Flux handles STT and turn taking + a pipeline for LLM and TTS managed based on Flux events"
X Link 2026-02-07T07:23Z [----] followers, [---] engagements
"day [--] learning to build voice agent infra from scratch: put together a VAD @twilio @DeepgramAI (SST) o4-mini @elevenlabsio (TTS) into an event-based loop that coordinates listening and speaking main issue right now is latency and quality of turn-taking/interruptions. not quite as good as Vapi/Elevenlabs off-the-shelf next I'm going to try using Deepgram's Flux for both STT and endpointing https://twitter.com/i/web/status/2019693047022289387 https://twitter.com/i/web/status/2019693047022289387"
X Link 2026-02-06T08:42Z [----] followers, [---] engagements
"just beat @Vapi_AI E2E latency with my own voice agent orchestrator. had to host everything just right to get the numbers this low Vapi's reported: 840ms mine: 690ms + 100ms (twilio) = 790ms that 50ms is a big deal in voice write-up incoming"
X Link 2026-02-09T09:21Z [----] followers, [---] engagements
Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing