Matthew Cosier

@coleywoleyyy lol. Your spec is ass.

the most detailed spec I've ever written

I don’t understand RT voice. Until the underlying execution model is at the same pace as the RT voice responder, it’s just poor UX. Cerebras where you at?

@reach_vb Fark. It’s only there on new threads… deerrrrrpp. My bad? Bad UX tho

ChatGPT Voice in desktop is now rolled out across all paid plans! Good time to give it a shot, ask it to kick off your /goals for the night ;)

@reach_vb I’ve updated and it’s still not there ? Must be missing something

@sama @jxnlco Waiting, waiting …

oh and also...750 token/sec coming to 5.6 sol in july!

Good new first: Sol is a smart, efficient, and a significant step forward. It is the same price as GPT-5.5. Also launching in the GPT-5.6 family is Terra, with 5.5-level performance at half the price. Bad news: at the request of the US government, it is launching today in limited preview instead of the open access launch we were planning on. We are working with the government to get to general availability as fast as we can. I think it is quite reasonable to roll out models--especially as they reach significant new levels of capability--in this way. It fits with our long-held strategy of iterative deployment. But this isn't quite the process that we think is optimal. Now we will with the government to attempt to get to a transparent, reliable process for early access, and to ensure that as long as our safeguards work as intended we can release widely. We want to be a reliable, dependable partner that works with all stakeholders, and we also want to live by our mission of benefiting all of humanity. I believe the government shares most of our goals, and that they are overall doing a good job in a very difficult situation. We will work as quickly as we can to get this model in your hands and we hope you will love it.

@reach_vb Still nothing in AU

voice in Codex is pretty wild! you can now literally talk to Codex while it works - kick off tasks, check progress, interrupt it, change direction, or start another thread without going back to typing feels especially useful when you’re figuring things out as you build excited for you all to experience it too ;)

@VictorTaelin Maybe you need to ask yourself why you built it to begin with. Did you build it to profit from it, or did you build it because you are passionate about it? Those are two very different things. They often don’t directly align. Linus didn’t do it for the $ either, but he did ok..

This is it! Bend is 99% ready. The 1% are problems I'm not aware of yet, but Fable will easily fix anyway. Except for one thing. I'm still not sure how to handle the business side of this... Opening it all OSS upfront is risky, it could kill HOC forever. Making the compiler closed source is a shame though, and may *cause* the project to fail. There is one application I have in mind, it is only possible with Bend, and it could *be* our business. But then I'd need to develop it before launch, which would take a month or so. I even considered going back to Kindelia, creating it in Bend, and letting Bend be nothing but PR for it...? But it sucks that we need to go through such extreme indirection just to make money from a programming language. That's the thing though, programming languages aren't really products and it really sucks that you can create the greatest language ever, make it go viral, have millions using it, produce a lot of real value, and still make $0. Is there even any other category of thing that is like that? I think the next week will be a bunch of Fables emulating users and proposing small improvements, while I just think about the business and how to handle all of this...

@thsottiaux Please let this be a model release. I don’t think so though. Prolly voice

Unbelievable excited for what’s coming together. Tomorrow is feeling codexy

Man OpenAI have flagged me 4 times today on cyber. Nothing of the sort being done - just basic bitch PR reviews. Must be a bug 🐛

@VictorTaelin @VitalikButerin lol should someone tell him.

@VitalikButerin bro

A new type of "high-level programming language" that seems really worth trying to make, is a language that gets compiled to Lean (or HOL, or...) that is specifically about making it as friendly as possible for a human to read definitions and theorems. Not the proofs - as all that matters with proofs is that the proofs are correct - just the definitions and theorems. The intended use case is that AI outputs a blob of proofs, and you're trying to make it as easy as possible for anyone reading the output to understand what the actual precise claims are that have been proven

@Mr_Salio Claimed 2m context window: check. Claimed it as a trust me bro leak: check. Claimed it to be surpassing fable: check. Calls everyone bro: check. Diagnosis complete.

🚨 Gemini 3.5 Pro Leak : It beats Claude Fable 5 > Gemini 3.5 Pro is reportedly place it ahead of GPT-5.6 Sol and Claude Fable 5 on several tasks > Gemini 3.5 Pro Still being tested with select partners > New Deep Think reasoning mode > Massive 2M token context window > Major improvements in coding, planning, and autonomous agents > Stronger frontend generation and SVG quality > Internal codename: Cappuccino Will Gemini 3.5 Pro put Google back on top of the AI race?

@bridgebench Uhm. This was on your post how it was the best thing ever lol. Aged well.

Kimi K3 is the slowest frontier model in the world. Makes it practically unusable.

@MarcosHernanz Going nowhere really fast 💨

I'm spending more tokens than the whole OpenCode team combined

@bridgemindai But it’s slow af? How can you be doing real work at those types of speeds… it’s just silly.

I just cancelled 2 $200 Claude Max subscriptions. In their place: one $199 MoonshotAI Vivace plan, purchased this morning for Kimi K3. Kimi K3 is better than Fable 5 at frontend design. That is half of why I kept 4 Max accounts, gone overnight. The other half is trust, and Anthropic burned it. Nobody knows if Fable 5 survives Sunday. Nobody knows what the limits are week to week. Nobody knows anything, because they refuse to communicate. Moonshot earned $199 with one great model. Anthropic lost $400 with a month of chaos.

@splitbycomma Did you get the followers you wanted ? Can everyone just bulk unfollow this muppet. Thx

You are NOT a real Software Engineer if you need to rely on the latest tech, e.g. LLMs, to get your job done. No serious company lets their engineers use AI.

Can someone hurry up and fire whoever it is at Anthropic who looks after customer comms. Thx

If you’re using line count to measure your efficiency you’ve got the wrong metric my boy. Try use customer centric metrics instead. Are you actually delivering proven value to the business and customers at higher cadence? Congrats, you won.🏅

10xing your output is not the same as 10xing the quality and throughput of your delivery. Going nowhere really fast comes to mind.

texting my gf like fable 5

Anthropic resets hit so much harder than Codex ones. Only because you barely ever need the Codex ones. Feeling fably today.