Joël Staub

Founder, AI, Claude Code, Codex

[Joël Staub @JoelStaub]

Ofc it's Opus 5 😂

[Péter Szilágyi @peter_szilagyi]

Presented without comments 🤡

[Abomination @Abomination81]

Astra is driving me fucking crazy. Literally have to tell it to finish something 5x to complete it.

Weird

[Parker Thayer @ParkerThayer]

This post looks like the start of a VERY sophisticated and well-funded PR operation to get support for Democrats to regulate AI into oblivion. Let me show you how it works:
1.) This guy, with minimal followers and no previous account activity, goes to the Wall Street Journal which publishes an exclusive with quotes from him on his resignation 18 minutes BEFORE this post goes up. Planning was clearly done in advance.

2.) Within hours, it has tens of thousands of reposts and the account has 100k+ followers. The post is punchy, quotable, it almost seems professionally written. The first three accounts to quote tweet it all do so within 15 minutes of the initial posting. Remember, this account had basically zero engagement beforehand, so an organic reach explanation seems unlikely.

According to Grok those accounts are @_NathanCalvin (General Counsel at Encode AI), @peterwildeford (Head of Policy at the AI Policy Network), and @DKokotajlo (Head of the AI Futures Project), all of which are up-and-coming AI-Doomer policy advocacy nonprofits.

The AI Futures Project website says it is funded “primarily” by the Survival and Flourishing Fund, which says on its own website that it has advised Jaan Tallinn, Skype creator and one of the leading investors in Anthropic, to grant over $2.5 million to the AI Futures Project since 2024.

Encode AI says on its website that it is ALSO funded by the Survival and Flourishing Fund, which in turn says that it told Anthropic investor Jaan Tallinn to grant $516,000 to Encode AI in 2025.

And wouldn’t you know it, the Survival and Flourishing Fund ALSO says it told Jaan Tallinn to grant $2 million to the AI Policy Institute, the 501(c)(3) affiliate of the AI Policy Network, as well.

What are the odds that the first three quote tweets of Coxon’s post would all be major AI-restriction policy advocates funded generously by the same donor, who also happens to be one of the leading investors in, and a board member of, Anthropic, the company Coxon was resigning from? And all within 15 minutes of posting (two within ten)?

3.) Jacob Coxon doesn’t have much of a resume, but we do know that, in 2022, he got a $20,159 scholarship for the “long term future scholarship program” from the Good Ventures Foundation, one of the philanthropic vehicles of Dustin Moskovitz, a notorious AI-doomer who has spent tens if not hundreds of millions on policy advocacy to strictly regulate AI, while also being an Anthropic Investor himself.

It also just so happens that the 14th person to quote Coxon’s post was @MaxNadeau_ (27 minutes after posting) who is the program officer for the Technical AI Safety team at Coefficient Giving, another of Moskovitz’s philanthropic spending vehicles. Max is not a frequent poster, his last posts before quoting Coxon were before Labor Day, but he was remarkably quick off the mark for this one.

4.) Basically every major Democrat politician and candidate has suddenly glommed on to this post, and conveniently, as the people cry out foe answers, Bernie Sanders already has a bill written to “ban super intelligence” and regulate AI into oblivion, and will be releasing later this week. The bill, among many other things, will create “a new cabinet-level federal agency to safeguard the public from the dangers of artificial intelligence” that will be “advised by an Artificial Intelligence Advisory Board comprised of experts on artificial intelligence.” Do you think, perhaps, Anthropic and its many investors who fund AI policy advocacy might have interest in getting to place a pet “expert” on the board of an entity that dictates what AI is and isn’t allowed to do? And isn’t it fortuitous that this whistleblower came forward with his oh-so scary stories so close in proximity to the release of the most radical piece of AI legislation ever introduced?

[camsoft2000 @camsoft2000]

Is it me or has Astra regressed compared to GPT-5.6 in regards to working autonomously? Without a goal it seems to stop after every small milestone and I have to keep telling it to continue.

[kitze 🛠️ tinkerer.club @thekitze]

it's not just me and its not "oh an addict got uses to the intelligence level".

it was fantastic on release day. literally 1 day later they crank the fucking slider down and it's an idiot and refuses to do things.

FUCK this shit.

[Marcin Krzyzanowski @krzyzanowskim]

gpt-6-astra xhigh is stupid beyond comprehension for my job. I literally question more than half of changes it made on the higher level, even before I look what it did on the lower level.

I tasked it with a specified job, that is not innovative, has prior similar work done, has established architecture etc. It worked few hours, and more I check, more I find what's broken. It ignored the architecture and made things unimaginable complex as a result. that complex part started to fall apart on each new thing build since. to fix edge cases it started to build on top without even trying to address the root cause I was able to scope in few minutes.

a definition of a slop happened 🤷‍♀️ and I feel violated.

[Irushi @Im_IrushiK]

Unpopular opinion: Something feels off with GPT-6 Astra.. May be??

Way more mistakes, way more apologies, and way more instructions needed just to get the same tasks done(I've literally experienced this). It genuinely feels like I’m doing more work even after using a powerful model. & This screenshot kinda explains the situation.

More repetitive work → more tokens → hitting usage limits faster → spending more money. Is this the plan, OpenAI?

[Joël Staub @JoelStaub]

It's the first time I had to take a step back with an LLM and change all my hermes cronjobs back to GPT-5.6 Sol, because GPT-6 Astra is just lazy af and really dumb in many cases (it then confirms it didn't read instructions/skills all the time).

And yes, I did spend many hours slimming down skills/crons/AGENTS.md etc.

I'm fully aware that Astra can be very smart, but it's truly hit and miss and thus, not really suitable as daily driver.

I do lots of system administration work.

[Peter Yang @petergyang]

After using GPT-6 Astra as my default model since launch, I’ve found it both incredible and frustrating to work with.

It’s incredible at building 3D games and models, high-level thinking and strategy, and computer and browser use.

But two things keep it from becoming my daily driver:

1. When I ask it to do something, it sometimes tells me how it would do it instead of just getting it done.

2. It doesn’t reliably follow the instructions in my AI skills.

The second issue is especially problematic because it keeps happening even after I cleaned up my skills. If anyone has found a solution, I’m all ears.

For now, I’ll stick with Sol for most execution tasks and use Astra for harder tasks or to get a second opinion.

I wrote more about my impressions of Astra in my new post: https://t.co/mcF7mGuVhL

[Joël Staub @JoelStaub]

Just fearmongering, period.

[Jacob Coxon @hilbertspaess]

I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.

[Joël Staub @JoelStaub]

The worst thing about Astra!

[Tyler @rezoundous]

Astra: "It will do A,B and C"
Me: "That is fine but I just want it to do A and B."
Astra: "Okay, it will do A and B."
Me: "...So have you done it?"
Astra: "No, would you like me to do it?"
Me: "yes wtf"

[Chris @ChrisGPT]

I’m sorry, but the $200 plan is basically unusable at these limits.

Running just 2 GPT-6 xHigh chats with 0 subagents can burn through the entire allowance in 2–3 days.

I know I’m not alone on this. Why not let us buy additional resets? Buying usage for GPT 6 is a joke, it will spend 5$ in 5 minutes. Forcing people to buy multiple subscriptions across multiple accounts is inefficient and unnecessarily cumbersome please let us buy weekly resets 🙏.

[Joël Staub @JoelStaub]

@GabeTognon @NousResearch 😂

[Joël Staub @JoelStaub]

@alex_prompter I'm sorry but Fable, sadly

[Robert Willacker @derWillacker]

Das ist bühnenreif.

[Joël Staub @JoelStaub]

Fully agree

[Theo - t3.gg @theo]

GPT-6 Astra has done incredible things I never thought a model could do. It has also done some of the stupidest things I've ever seen a model do.

Generally speaking, Fable 5.1 just does what I ask.

[Joël Staub @JoelStaub]

I think I never raged that much against an AI than with GPT-6 Astra (medium effort), using Codex and Hermes.

The model constantly stops/pauses

And yes, I cleaned up all instructions, skills thoroughly etc. according to official OpenAI dev docs.

It doesn't even do the simplest tasks right without correcting it, like writing a normally understandable email, like leading what an email is about first!

You literally have to babysit it for simple daily tasks.
It can be smart sometimes, yes, but mostly it's pretty annoying to work with!

@robinebers I think you're sort of in the same boat, no?

[Robin Ebers @robinebers]

going through a major Astra crash out

[Alex Dochioiu @AlexDochioiu]

It hurts me to say this, but Claude 20x includes way more Fable usage compared to Astra in GPT 20x.

This just sucks.

[Joël Staub @JoelStaub]

@turkelitopcusu @argofowl Explicitly for subs too, confirmed by erich from openAI

[Joël Staub @JoelStaub]

@robinebers I talk the same to Astra, damn it's frustrating!