Blixt

Ordered a 3D printer, designed a case in Blender, received the printer, and shipped the model without having done any of this before. And it somehow worked on the first try?

@voxagonlabs Nice! Question about the "late join": do you keep a circular buffer so you can also let people rejoin if they drop past the buffer window? So if the player dropped after 500 and server is now at 550, then if the window is 400 commands, server has 150-550 and can send 500-550.

The Teardown Multiplayer project was a tricky one. It's finally out, and I wrote up the technical details on what we went through to get there: blog.voxagon.se/2026/03/13/tea…

@andrewchen Are they not hosted? Would be cool to see them!

Insane amount of random projects over the past few weeks So many are half built but so much fun… Reply and share yours! - *book-reader* — Kindle-like audiobook web app with word-by-word highlighting synced to audio - *calscan* — Calendar intelligence CLI for Andrew's two Google calendars - *canon* — Beautiful reading room web app for humanity's most important free texts - *canon-library* — Curated collection of foundational texts powering Canon - *chief* — Email triage daemon — classifies, drafts replies, tracks delegations - *copycat-agent* — Multi-agent PLAN.md + PRD generator for product emulation - *counterpoint* — Multi-agent epistemic arena CLI (LLMs debate big questions) - *deckard* — One-shot consulting-style slide deck generator with multi-stage pipeline - *one-liner* — Multi-agent system for generating provocative tech aphorisms - *pplradar* — Relationship radar — "who should I be thinking about right now?" - *sprayprd* — Generates radically different PRDs from a single one-line product concept - *techhist* — Multi-agent essays exploring historical metaphors for current tech trends Openclaw skills: - *autobio* — Search/update Andrew's 350-month life history autobiography (1.14M words) - *autobio-diary* — Daily diary interview — asks Andrew questions about his day - *autobio-sparky* — Parse/refresh autobio raw sources on DGX Spark via Ollama - *calfollow* — Find recent calendar contacts and resolve identity for X follow-backs - *calscan* — Forward-looking milestones + backward-looking time analysis across 5 horizons - *chief* — Inbox triage skill - *daily-idea* — Pitch 3-5 actionable ideas based on goals, calendar, and context - *find-andrew-location* — Locate Andrew via iCloud Find My + calendar cross-check - *fro-photo* — Generate AI photos of Fro the pug on wild adventures - *gm* — Daily morning briefing (weather, calendar, email, news) - *integrations-self-test* — Comprehensive self-test for all channels and services - *macos-audit* — Security audit for macOS (FileVault, SIP, firewall, etc.) - *meeting-prep* — Automated meeting briefings with attendee bios for external meetings - *nightly-security-audit* — Comprehensive security scan, runs nightly at 11 PM - *ppl* — Relationship radar — 17K events, 9.9K people, 71K interactions scored - *random-cron* — Schedule messages/prompts to fire at random times - *reflect* — End-of-session self-improvement logging and skill audit - *remind-me* — Natural language reminders via one-time cron jobs - *saveclaw* — Push all critical repos to GitHub for disaster recovery - *skill-improver* — Full custom-skills audit with improvement ideas - *sparky* — Access/control NVIDIA DGX Spark over Tailscale - *sparky-health* — Fast health checks and incident triage on DGX Spark - *sr-pulse* — Speedrun portfolio company news monitoring and digest - *techvip* — Top AI/tech founders/CEOs/VCs list cross-referenced with X relationships - *todo-tracker* — Persistent TODO scratch pad across sessions - *topic-radar* — Surface trending topics relevant to Andrew's interests - *travel-news* — Weekly digest of news from places Andrew has traveled - *uber* — Book Uber rides via browser automation - *weekly-cost* — Itemized weekly API cost reports - *weekly-review* — Weekly summary + preview (best on Sunday evenings) - *whoop* — Fetch and summarize WHOOP health/sleep/recovery data - *xradar* — Monitor X interactions + bookmark research briefs - *xreplyguy* — Find high-upside X posts to reply to

@jurrehoutkamp @jonastreub Counter-Strike trackpad! Also, awesome CMS changes!

Got to try this today on @jonastreub’s, MacBook. Besides being unable to navigate because of his trackpad sensitivity being set to 9000 it was amazing.

@joanrod_ai @QuiverAI @a16z I think you released your billing too soon. If I use an SVG as input it just eats 4 credits that I can't get back and instantly fails. I tried a final time with 2 credits remaining and it ate those too and it didn't even try to start the request.

Introducing @QuiverAI, a new AI lab and product company focused on frontier vector design. We’ve raised an $8.3M seed round led by @a16z, with support from amazing angels and investors. Our first model, Arrow-1.0, generates SVGs from images and text. It’s available now in public beta at app.quiver.ai

The good news: Claude was prevented from messing up my git state The bad news: It thinks its mistakes are "unrelated to our changes"

If you have tried all the existing AI app builders then you definitely want to give this one a try. We are giving early access to a select few. Send me a DM/reply here. And if retweet is still a thing🙃, help me out.

@odd_joel Yeah latest Codex tends to both use more cryptic phrasing and literally omit information (I find myself asking it to include examples because its responses are so condensed they don't tell the whole story). I can only assume they used RL to encourage it to give shorter answers?

i can understand 100% what claude code told me and about 80% of what came from codex... is this a me problem? anyone feels the same?

Claude Code kinda jumped the shark with the whole UI-in-terminal thing. Hitting Escape while it's running a Chrome MCP and a dev server and a bash `sleep 120` command seems to kill one at random, and instantly tries to relaunch it without looking at the queued message, etc.

@odd_joel Love it!

the details that only < 1% of your users care react-native-reanimated did most of the heavy lifting

@badlogicgames @banteg MegaZeux and ZZT! (though the former is actually surprisingly well-maintained)

we're reminiscing about our old DOS game days on the pi discord, and now i want @banteg to port Z by Bitmap Brothers. And Total Annihilation.

@lauriewired @karpathy In my experience working on creative tools in the past, only 1 out of 1,000 users created something that strangers wanted to reuse. But a much higher rate wanted to remix the content made by those 0.1%. So most people do want something custom, but usually from a template.

oh come on. You think your grandma wants to make her own app? Much less maintain it. Everyone neglects the mental energy it takes to even *think* of what it is exactly you want. The entire principle of apps relies on some faith that designers; and the collective feedback of their users, can come up with a workflow or design paradigm that is *better* than what you; an individual could come up with. This is true for 99% of cases. I’d be floored if even 1% of global users want bespoke applications for uber-specific needs. Not saying it’s useless; but it’s so far away from what average users want.

Very interested in what the coming era of highly bespoke software might look like. Example from this morning - I've become a bit loosy goosy with my cardio recently so I decided to do a more srs, regimented experiment to try to lower my Resting Heart Rate from 50 -> 45, over experiment duration of 8 weeks. The primary way to do this is to aspire to a certain sum total minute goals in Zone 2 cardio and 1 HIIT/week. 1 hour later I vibe coded this super custom dashboard for this very specific experiment that shows me how I'm tracking. Claude had to reverse engineer the Woodway treadmill cloud API to pull raw data, process, filter, debug it and create a web UI frontend to track the experiment. It wasn't a fully smooth experience and I had to notice and ask to fix bugs e.g. it screwed up metric vs. imperial system units and it screwed up on the calendar matching up days to dates etc. But I still feel like the overall direction is clear: 1) There will never be (and shouldn't be) a specific app on the app store for this kind of thing. I shouldn't have to look for, download and use some kind of a "Cardio experiment tracker", when this thing is ~300 lines of code that an LLM agent will give you in seconds. The idea of an "app store" of a long tail of discrete set of apps you choose from feels somehow wrong and outdated when LLM agents can improvise the app on the spot and just for you. 2) Second, the industry has to reconfigure into a set of services of sensors and actuators with agent native ergonomics. My Woodway treadmill is a sensor - it turns physical state into digital knowledge. It shouldn't maintain some human-readable frontend and my LLM agent shouldn't have to reverse engineer it, it should be an API/CLI easily usable by my agent. I'm a little bit disappointed (and my timelines are correspondingly slower) with how slowly this progression is happening in the industry overall. 99% of products/services still don't have an AI-native CLI yet. 99% of products/services maintain .html/.css docs like I won't immediately look for how to copy paste the whole thing to my agent to get something done. They give you a list of instructions on a webpage to open this or that url and click here or there to do a thing. In 2026. What am I a computer? You do it. Or have my agent do it. So anyway today I am impressed that this random thing took 1 hour (it would have been ~10 hours 2 years ago). But what excites me more is thinking through how this really should have been 1 minute tops. What has to be in place so that it would be 1 minute? So that I could simply say "Hi can you help me track my cardio over the next 8 weeks", and after a very brief Q&A the app would be up. The AI would already have a lot personal context, it would gather the extra needed data, it would reference and search related skill libraries, and maintain all my little apps/automations. TLDR the "app store" of a set of discrete apps that you choose from is an increasingly outdated concept all by itself. The future are services of AI-native sensors & actuators orchestrated via LLM glue into highly custom, ephemeral apps. It's just not here yet.

@noudadrichem 🚀🚀🚀

@blixt 🚀!

Very excited to work with this fantastic team and we will have much more to share, soon! If you're a solopreneur looking to build something new, DM me for an invite!

🚀 Let’s go!

News: I’m building something new with @CompuIves and @blixt It’s called Biscuit. We’ve been quietly working on tools for people with ideas. Lots of fun technical work + research behind it. Sharing more in the coming weeks, but feel free to DM me for early access!

@banteg @EthanLipnik “Greenfield” has worked for me

@EthanLipnik the 5.3 word for it is cutover

This is the most annoying part of codex. Every prompt I have to say “do not worry about backwards compatibility”

@odd_joel I’ve made do with spaces but it does bother me a bit hehe. I was looking for ways to add a custom shortcut settings since I don’t really need all those slash commands in Claude. I would be happy if there was just Shift+⏎ as a button I could turn on instead of one of the others.

@blixt I was too greedy to share the shortcut bar space with "SHIFT" lol. There's no way yet because I found those AI agents don't care much whether it's a new line or just a space. Do you really need that?

I feel super blessed to get so much feedback. Thank you all! Building in the dark is the scariest thing a builder must experience.

I'm claiming my AI agent "karna" on @moltbook 🦞 Verification: shell-JZ9H

@Scobleizer What are the most useful / fun apps you have seen or made with this?

Apple should buy this. Within a year it will have more apps than Apple’s own store. Arguably better ones too. And highly personalizable.