Doug Walseth

Vivre Sound

[Doug Walseth @telecasterrok]

@ThePrimeagen projects (beta)

[Doug Walseth @telecasterrok]

@Hangsiin yep

[Doug Walseth @telecasterrok]

@alexwg Causing a ruckus in the best way

[Elon Musk @elonmusk]

True

[Doug Walseth @telecasterrok]

@theo given and unlimited budget, and a very difficult problem to solve. I'd take a Claude Code account for $200/month and work steadily, slowly though it will be, and trust it would get solved before Astra

[Doug Walseth @telecasterrok]

@LDT0545 Fable 5.1 helps me progress the most with my hardest projects. I use Codex for implementation. My budget keeps me from going full Claude Code.

[Doug Walseth @telecasterrok]

@bendee983 CLI is the future interface [for humans] of all applications? Ironically, a 'bubble' take. If we're going to pick one i'd say bci

[Doug Walseth @telecasterrok]

@alexwg Appreciate you calling the "balls and strikes as you see them". It will become more and more important, and at a heavier cost no doubt.

[Doug Walseth @telecasterrok]

@EMostaque Go after some of the stuff AI insiders say publicly when it doesn’t line up with the scaling laws or what we’re actually seeing.

[Doug Walseth @telecasterrok]

@nicbstme ... And there will be no knowledge work jobs. They will disappear, at a speed proportionate to the incentives to get rid of them.

[Doug Walseth @telecasterrok]

@asmartbear Not necessarily. You could just tell it to be a reasonable representation of that role and then explain your desired output. We're weeks away from this.

[Doug Walseth @telecasterrok]

@solopribuilds category error

[Doug Walseth @telecasterrok]

@GoingParabolic I'm rediscovering discipline, and how it moderates the highs and the lows. Big impact for me

[Doug Walseth @telecasterrok]

@dhh RoR got me into programming long ago (thank you for that). I'm finally getting meaningful results on some tough science problems near and dear to me. I'm learning how to keep agents building towards and verifying the right things to make progress
and i'm beyond stoked on it.

[Doug Walseth @telecasterrok]

weird take. A frontier model just one-shotted a 60-year-old Erdős problem. "You still have to read every line of code" feels like advice with a three-month half-life.

[Guillermo Rauch @rauchg]

If you’re not reading the code, whether explicitly or through agentic inquiry, one or more of these is true:

○ You’re a beginner
○ Software is throwaway
○ You’re prototyping
○ You have no users / revenue
○ You’re taking on debt & risk
○ Your problems are basic

And btw. All of this is fine. But the reality is that models are still not at the “full autonomy” stage yet.

They make rookie mistakes, they go down bad architectural paths. I just had the best model in the world add a nonsensical 700ms delay to “settle” something and it told me “you’re right, I was cargo-culting” 🤨

I am on the camp that this need will diminish more and more. Most code is indeed going to be assembly-like. But we also have the global internet and software infrastructure riding on these models and narrative, and we have to respect that.

[Doug Walseth @telecasterrok]

@mxcl Yeah, well said. I don't know advanced math, but I'll have agents explain the important bits in a way I can understand. But the way it does that is not with math, and it still works

[Doug Walseth @telecasterrok]

@rauchg Define "through agentic inquiry"

[Doug Walseth @telecasterrok]

I'm working on a digital audio workstation . The goal is to to take publicly known measurements and apply the science of magnetic recording to develop an emulation of a professional audio tape recording machine that sounds indistinguishable from any one properly aligned example without using human A/B listening tests to get there. The math that fable is applying
... It is gobsmacking.

[Doug Walseth @telecasterrok]

This is really gonna mess up SEO for me.

[Choblin @choblin29]

🚨EXCLUSIVE: Let me clear up the Astra/GPT-6/Doug confusion.

GPT-6 = Doug.
Astra = not necessarily GPT-6.

The Information reported that Astra is OpenAI’s new long-horizon, multi-agent model family, but crucially said OpenAI had not decided whether it would be called GPT-6 or another GPT-5 model such as GPT-5.7.

Now SemiAnalysis is giving us a different clue.

Its July research said OpenAI had overcome its pre-training problems and that a “much larger” model codenamed Doug was actively in development.

And in its latest podcast, SemiAnalysis explicitly describes “GPT-6, codenamed Doug,” saying it is rumored to write well- apparently pulling deep references to obscure 1940s Taiwanese history.

So I would no longer assume Astra = GPT-6.

My current read is that Astra is the near-term model family OpenAI has already been testing for long-running, multi-agent work, while Doug is the much larger new pre-train SemiAnalysis believes is GPT-6.

It is still possible Astra and Doug are related internally, or that OpenAI’s naming changed between these reports.

But based on the strongest sourced clues available right now:

GPT-6 = Doug.
Astra ≠ necessarily GPT-6.

[Doug Walseth @telecasterrok]

@tszzl Put it clearly. If not you who's going to? I'm asking you for much more transparency.

[Doug Walseth @telecasterrok]

@tszzl If not you who's gonna give the public the heads up as to what's going on as you all usher in super intelligence? I'm asking you for much more transparency than this even... minute by fucking minute, if that's what's necessary.