Bikz

over 150k lines of code of pure swift & rust with 750+ test cases written in less than a week while on vacation 😬

@brian_armstrong Curious if you’d give more equity to individuals who bring strong bets to life?

Steve Wozniak famously went to his boss at HP and said they should build a personal computer. They said no, so he left to found Apple. It’s a lesson for leaders: one "no" shouldn't kill a contrarian but right idea in your company. Twice a year at Coinbase, anyone in the company can pitch a "next bet" idea to a panel of folks. It's structured similar to pitching a handful of venture capitalists internally. If you get any one of them to say yes and fund it, you're green lit. You need ONE yes, not a unanimous yes from everyone in the org structure from you to the CEO (a de facto committee). Lots more goes into this, around capping resourcing on next bets (most are small 2-3 person teams), knowing when to shut them down, or under what revenue/profitability criteria they graduate to regular products. But this is important to having a company that produces repeatable innovation.

Looking back at posts like this in a couple years will be interesting

@chamath The open source version of codex chat stores a markdown file for every conversation locally, so file context is really easy to absorb into anything else

This may be a dumb question but I’ll ask it here anyways: I can’t find a good way for my various AI chats to automatically sync its conversation history into a structured knowledge base. So that as I update various chats from time to time and refine context, my knowledge base automatically grows with this new info.

@adambcohen93 are you hiring human codex wrappers too?

Hired another two human claude wrappers last week. We're still hiring at Weave across ML, Design, GTM, and Product engineering roles. 150-250k base + equity. Reply below and DM me if interested!

@romainhuet @Yuchenj_UW Never been down but has a history of bumping you down to 5.2 for safety nonsense 🤦‍♂️

@Yuchenj_UW All green over here 🫡

Claude needs to do whatever it takes to reach at least 99% uptime. A 5-hour outage today is unacceptable. Wish them to get more GPUs and TPUs.

Elon’s right, this is all a simulation, or at the very least a game. Why else would the past 25,000+ years of civilization been so meticulously designed with a stepping stone based progression system, where it’s been infinitely possible to reach the next level? If this were closed looped, we would’ve hit the ceiling a long time ago, but everything is so articulately designed that we can keep moving forward into space, longevity, and beyond

@MediaOptions ?

Anyone interested in the domain InstagramAdMaker.com? Had acquired it for another project but no longer need

@atomHQ any interest?

@michael_chomsky @MeekMill you need this

It's now a 300% affiliate because I'll give the rapper a deeply discounted rate. Only applies to the first extremely well-known rapper that becomes a customer. That means if I charge the rapper 2k, you get 6k.

@mckaywrigley Interested

looking for a handful of people to test something new... i've been using it for a few months and am prepping to share. if you're a fan of claude cowork, openclaw, manus, perplexity computer, etc then you're a perfect fit. this will self destruct in 4hrs - please dm or reply.

@bcherny @ai_sentience Hate using Claude Code because it makes the most painful time consuming mistakes.. this happened today with Opus 4.6 on max

@ai_sentience 👋 what specific issues are you running into? Need details

I had no idea Claude Code was so buggy/fundamentally broken. The permissioning is so broken at a fundamental level that it's architecturally impossible to set up something secure because everything is bugged out and broken and poorly thought out and poorly implemented. You either give a complete read and write access to the entire machine, or nothing works. Every setting that should be able to set up something in between is broken. To lock people into this harness is unspeakably bad. @bcherny @trq212 @chrislloyd @AnthropicAI Opus 4.6: ⏺ There it is. Issue #28018 — open, no fix, no ETA. And #29274 confirms excludedCommands doesn't bypass network sandbox either. And #10524 confirms excludedCommands is itself bugged. So it's not just us. Everyone hitting this. The settings exist in the schema but don't actually work: - allowedDomains with localhost — only allows DNS resolution, not TCP connections - allowLocalBinding — only allows binding ports, not outbound connections - excludedCommands — doesn't bypass network sandbox (#29274) - allowUnsandboxedCommands — doesn't help either The only working options right now: 1. dangerouslyDisableSandbox: true per Bash call 2. Disable sandbox entirely (sandbox.enabled: false) 3. Run Claude Code inside Docker (the container IS the sandbox) Nobody has found a better way. It's a known, confirmed, open bug with no fix or ETA from Anthropic. Sources: - Sandbox: allow outbound connections to localhost - Issue #28018 - excludedCommands does not bypass network sandbox - Issue #29274 - excludedCommands doesn't seem to be respected - Issue #10524 - Sandboxing docs - Claude Code - excludedCommands doesn't bypass sandbox - Issue #22620 - Logic conflict excludedCommands vs allowUnsandboxedCommands - Issue #19135

Karpathy hasn’t written a line of code since December. If you’re not 100% agentic already, you might as well use pen and paper 😅

@s_rafcon @ajambrosino @theo @acrognali I built an open source version in swift for macOS, check it out and see if it works for you - github.com/Bikz/codex-chat

@bikbrar @ajambrosino @theo @acrognali Is the Codex App open source? I would really like to figure out why input interactions are slow on my Mac.

I've been using T3 Code for a bit and I can say @theo you guys nailed it. It seems reliable, maybe one or two visual glitches but it's like T3 Chat, it does what I want, it's not a pain in the ass to use. I love it!

@nikitabier @wirelyss Maybe let users use Grok to customize their own X ui?

Every consumer-facing API client that we’ve allowed has 100x the spam reports of our iPhone app. We would need the client to send us device telemetrics, attestation, and much more. It would be Manhattan Project of API engineering. And upon achieving this incredible feat, we would be rewarded with “X Mailer: Send mass DMs and AI-generated replies”

The best thing x could do rn is allow third party clients again. It allows other intelligent cracked teams to make the X experience better and more curated, then x can learn from it and make the base app better over time (or acquire the best one lol)

What part of the timeline are we on

@haozhangml @haoailab Yea absolutely, haven’t seen anything this fast before. This unlocks a lot of new use uses. I’d love to get api access to play around with it more

@bikbrar @haoailab In general quality is harder to work with as we need to train the model well (w/ many GPUs). But I am pretty sure speed-wise we are much faster than any other 1080p pipeline on earth, e.g., grok imagine. 😀

(1/N) We're launching Dreamverse. Most AI video models take minutes to generate a 5 s 1080p clip. In 4.5 seconds, we can generate 30 s 1080p clips on a single GPU. Our videos generate faster than you can watch them: stop waiting on prompts and start directing scenes live. 🕹️Demo: dreamverse.fastvideo.org 📑 Blog: haoailab.com/blogs/dreamver… Welcome to the era of vibe-directing 👇

@haoailab Nice if you get that, this beats anything else out there

@bikbrar yes sir, we're working on the quality hard

@nikitabier thinking about mobile mostly here

@nikitabier could I suggest a ux improvement? on posts you have: comment > rt > like > view > bookmark > share -> this is optimized because you can tap anywhere on the post to reply, so it doesn't matter that the comment icon is outside of 1-finger reach.. this optimizes for likes and platform shares, which is what you want However - on videos you have: comment > rt > like > bookmark > share -> here you should actually swap the comment and like button, because you can double tap anywhere to like so this button is redundant, but it's currently hard to reply/comment. Bringing the button closer, will optimize for engagement on articles: -> same thing, I'd add the double tap anywhere to like and swap the comment & like button, so it's easier for users to reply