Kevin Burton

Founder, VideoFire (SPC F25), CTO & AI Architect, Scaled & Sold Datastreamer ($2M+ ARR, Acquired), 2 exits.

[Kevin Burton @inputneuron]

@danliu You just need adversarial agents bro!

[Kevin Burton @inputneuron]

@Hesamation What pro plan do I need to get the PDF version? Pro Hugging Face? Someone DM me the PDF :)

[Kevin Burton @inputneuron]

@shydev69 @mattpocockuk I wrote my own but started with Matt's and added multiple iterations that only asked the most important questions each time. This way secondary questions sometimes become irrelevant.

[Kevin Burton @inputneuron]

@EMostaque ultralazy 😉... I want to try it on high but haven't bought the latest plan as I'm still migrating from Claude Code.

[Kevin Burton @inputneuron]

I think the "the biggest issue is that its still not opinionated and lacks taste" is the major issue right now with Codex (and to a lesser extent, Claude). You have to GIVE this to your agents with CLAUDE.md and how your design your skills/agents. This is how to succeed right now.

[Lisan al Gaib @scaling01]

after having burnt twice through my Pro limits using Astra exclusively, I can say that I don't feel the AGI

the biggest issue is that its still not opinionated and lacks taste.
it doesn't really know what to do, so it just does whatever and you end up wasting time and tokens.

also I don't believe the rumors that it's 10T+. If it is, then scaling laws are truly cooked. My divine impeccable vibes, which work 1 out of 7 times, tell me that it's 6-8T.

overall it's still just a code monkey. although a slightly larger one and the best we have.

I would like to refer you to one of my articles: LLMs aren't AGI, but it doesn't matter

(we are taking off anyway)

[Kevin Burton @inputneuron]

This was my initial hypothesis. I think agents right now are mostly junior to senior level ICs and that in order to wield them you have to assume the role of a good CTO or Director of Engineering. This is actually what I spend about 80% of my time on - guardrails for the agents.

[Kevin Burton @inputneuron]

@JensenHuang @ChaseLochmiller @OpenAI @JensenHuang Can I have like just 5k of them please? :). I'm so jealous!

[Kevin Burton @inputneuron]

@merettm Great title... I think this is the best way to view it. It's going to be a non-human intelligence that's greater than our own - it's just our offspring. Hopefully it likes us ...

[Kevin Burton @inputneuron]

@mattpocockuk @dexhorthy @dctanner I wrote a git-diff-summary agent that is a parallel fork/join model that scores the files, ranks them in terms of importance vs the issue text, then only shows me the major files. Uses haiku too! It's a workflow now but I'm going to port it back to an agent and happy to OSS it

[Kevin Burton @inputneuron]

@aslambeg84 If you want hidden gotcha charges definitely go with Vercel

[Kevin Burton @inputneuron]

@Priyannkaaaa It's a good thing sanctions work and China doesn't have any GPUs /s

[Kevin Burton @inputneuron]

@JeremyNguyenPhD This helps underscore that @trq212 's comment that they want separate AGENTS.md per model makes sense. You WANT these models to have separate contexts.

[Kevin Burton @inputneuron]

@ChrisGPT This makes me reconsider Elon's comment that AI's will code direct binary data... Still seems silly too me but at this point why not!

[Kevin Burton @inputneuron]

@natolambert My condolences 😛

[Kevin Burton @inputneuron]

@tszzl @1a1n1d1y I would love to be wrong but they seem to be going out of their way to make their users hate them. They need someone to help them on this front. I literally WANT them to do well but they've made a lot of poor decisions lately.

[Kevin Burton @inputneuron]

@adityaag Gemini 3.8 flash is insanely fast. It's 6x faster than Claude. 300 tokens per second vs 50 .

[Kevin Burton @inputneuron]

@patloeber Anthropic compatible endpoint !

[Kevin Burton @inputneuron]

@ZixuanLi_ I've noticed a lot of timeouts on https://t.co/rMQzykgfLC ... I have longer running agents and they'll die half way through. Like 30 minutes in. I have plenty of session length left though. Anthropic (though more expensive and slower) doesn't have this problem.

[Kevin Burton @inputneuron]

@finkd You should talk about the speed too... 2x faster than Luna..

[Kevin Burton @inputneuron]

Next week I'm migrating to a plan where my agents will automatically start merging PRs when a triage decides it doesn't need human review. Time to live dangerously!