Rob @systemwithin

@r_marked I had this same issue and resolved it by putting this at the top of my global AGENTS.md Sounds dumb but it works. Another added bonus I noticed was that Astra is much more willing to work autonomously over long periods if I tell it to 😂

@matteing I think I found your Silicon Valley archetype

@signulll I sent you nothing, so I guess I can say, you're welcome 🌚

@thexanaxlady Niche? or maybe just not an algo'ner 🌚

@levelsio Happens to me all the time, I hate it

My lord. This man gets it 😂

alexei @alexeixbt

4 days ago

this guy literally explained the truth about the "ADHD superpower"

578 6K 40K 1.6M 20K

🚨 Codex Users 🚨 If you use a "Project Manager" agent to orchestrate tasks across workers in other Work-side threads, I highly recommend setting up ChatApp-side "Support Ticket" threads for specific use cases. These have been an insane unlock for giving my Project Manager agent a third-party-esque "consultant" it can turn to for grounded responses when it hits problems that would otherwise cause looping. It has also prevented a ton of false-positive "Blocker" stoppages that would have otherwise terminated the session. Moreover, it has saved me a ton of tokens by preventing agents from looping during long sessions (4-12+ hours) where I'm not always in the loop. Lastly, if you're like me and still iterate on a project from the ChatApp side, you can put that "Support Ticket" thread directly inside the corresponding ChatApp project so the LLM has access to whatever project source files you give it. Hope this helps you as much as it helped me!

@dhh We're focusing on the wrong statistic. x.com/systemwithin/s…

Rob @systemwithin

3 days ago

The more interesting statistic is how many Americans think AI won’t benefit the country because Chinese psyops are apparently working overtime.🤫 Because trust is a very delicate thing: once it’s gone, it usually stays gone. The way to win people back isn’t by glorifying AI or

0 0 0 61 0

The more interesting statistic is how many Americans think AI won’t benefit the country because Chinese psyops are apparently working overtime.🤫 Because trust is a very delicate thing: once it’s gone, it usually stays gone. The way to win people back isn’t by glorifying AI or endlessly preaching its benefits and breakthroughs. If anything, that can exacerbate the problem when there’s already so much “capability-doomer-shit” circulating. The solution is building infrastructure that preserves source history, so claims can be traced back to their origin and actually vetted. If people can understand that their “strong belief” may have been shaped by information presented in bad faith, then maybe we have a fighting chance. Until that changes, and until awareness of this problem reaches the mainstream, the divisions will only deepen. And the AI industry is far from the only victim.

Nick Gillespie @nickgillespie

4 days ago

'An average of 72 percent of respondents said they felt curious, happy or excited about A.I., compared with 41 percent who felt worried, sad or angry.' @nytimes's Damien Cave uses @Gallup data to remind us the rest of the world digs AI. (Annoyingly, NYT doesn't link to Gallup

38 76 384 356K 149

Yeah, I should have clarified: AI is the accelerant rather than the original cause. My point is that accelerants change the consequence profile. People tolerate abstract privacy risk because the harm usually feels distant and fragmented. I think the line gets drawn when accumulated data exposure produces a visible failure large enough that ordinary people can connect “all this stored data” to a concrete consequence. We're basically, living a real life "Tale of the Leaky Roof." I hope I’m wrong, but unless we make significant changes quickly, I can see something like “The Great Breach” happening: an unprecedented data exposure event unlike anything we’ve witnessed before. That’s why I used the fire metaphor. The dead brush was already there; AI just makes ignition easier and the burn faster.

I’m using it stand-alone right now, but with native Codex plugins. Huge QoL too: you can prompt your agent to make basically any plugin you want. Not trying to exacerbate your FOMO, but I genuinely think once you start using Codex, you’ll have a hard time going back to other IDEs, or even Claude Code because the computer use is excellent. I can show you my setup sometime over Discord or Google Meet if you’re interested in switching. I run a project manager thread that orchestrates across my other threads. I do miss VS Code sometimes though 😢

I don’t think the pushback from non-tech people / potential customers will be formal. I doubt most users will organize around “human-verified” as a category. It’ll probably look more like: person launches thing, no one uses thing. The pushback I do expect in the near trajectory is around data privacy and stronger consumer protections. If this current wave of AI teaches people anything, I think it’ll be that we made a massive mistake over the last 20 years by allowing nearly every digital company we interact with to collect and retain pieces of our digital fingerprint. I think the current surface area of data exposure looks a lot like fine fuels accumulating in a forest. Most people are still staring at the live trees. I think it may take one devastating digital fire before everyone stops focusing on what’s visible above ground and starts paying attention to the dead brush underneath. And I say this as someone very pro-tech / AI / future, so I don’t mean this as doomerism.

@amritwt 2026 tamagotchi 😂

@zenorocha GPT recommended your service to me for my project It was not its default recommendation, but after I requested further research, it found you guys offered the best solution for what I wanted

@dhh I sincerely appreciate your optimistic projections and takes about what the future holds for builders 🙌🏽

@TylerCLaprade @ericzakariasson I understand what the prompt does. My comment was more targeted at this process: → AI, design me an agent harness → AI, this harness is complicated, fix it → AI, optimize the optimized harness At some point, you need to own the shape of the system yourself.

@sarada_io I’m not sure how teams or orgs officially run their agents, but based on what I know about humans, I doubt it

The usage and quality-consistency struggles littered across the timeline make something abundantly clear: Anyone who wants to master agents needs to hard pivot toward environment engineering. x.com/i/article/2102…

@alexeixbt 😂😂😂