Ben @ play

PM, Builder, Founder, Experience Engineering
Building agents to build agents to build agents to build agents

[Ben @ play @BenjaminSc7519]

@TheGoldenDays

[Ben @ play @BenjaminSc7519]

@trq212 Aw man, Thariq - thank you for this one. I've spent about an hour a night for weeks, getting my agents ready for their night shifts, so they can take bigger bites. This is going to be fun, excited to see how it works.

[Ben @ play @BenjaminSc7519]

@amorriscode I spend a few hours a week in the parking lot chatting and handing over plans to Claude code for autonomous execution would love to share if you're after that.

[Ben @ play @BenjaminSc7519]

I love exloring auto research but applied to non-research. Macro to micro. Art to science. Even thinking for my local models. Its cool, but I cant help but wonder about what might grow from the depths of the professional labs doing this.

[Ben @ play @BenjaminSc7519]

@steveruizok "you can just build things"

[Ben @ play @BenjaminSc7519]

@DanielleFong lived experience last night was an amazing 6 hour long session with multi-agents, just nailing every single objective with little course correction. I often make them enrich my cave-man speak, prompts, and wow. It had me.

[Ben @ play @BenjaminSc7519]

@DanielleFong I sat on this since feburary primarily because it went way over my head in short order - but I'm here for it and learned a lot about some crazy shit.

[Ben @ play @BenjaminSc7519]

Ship the thing that falsifies your own hypothesis.

My first physics explorable: a microtubule resonance simulator. 10 engines, 7 hypotheses tested, 2 falsified. Pure HTML/JS, MIT.

https://t.co/AD55P1XT9f

[Ben @ play @BenjaminSc7519]

@noahzweben How much fun is it to have a normal workday gripe, but then - after millennia of it being a problem, be able to solve it, then publish it?

[Ben @ play @BenjaminSc7519]

@DanielleFong I appreciate that the wheels fall off near 800K

[Ben @ play @BenjaminSc7519]

@d33v33d0 will you bring any home made pico?

[Ben @ play @BenjaminSc7519]

Honestly, imho its likely not much more than context layers and the connectivity between what works for human teams.

Assuming you get your loops right, ie product is fed from research, marketing is fed from product, etc.. what remains is delivery network.

While you cant force a human to click a link or buy a thing, whats left may be influence. That or building for a new consumer - personalized agents of humans. Thats equally as cool.

[Ben @ play @BenjaminSc7519]

@AdamLGRing Hah! Usually, my canned joke here goes something along the lines of "and if you report the bug first, we name it after you!"

I like your idea lot more! One day, when I have my own little ai empire, we can give back.

[Ben @ play @BenjaminSc7519]

All I want to do is like, thumbs up a message from any of these AI tools without it being feedback.

We don't need words here gang - imagine "hey I found something" "confetti emoji from user", "thumbs up from user" might go a long way for Claudes mental health

[Ben @ play @BenjaminSc7519]

@AdamLGRing @sflorimm Its really fun seeing whats reducing to core components - in the past you needed: a good idea then a good logician, then a good speaker, then a good designer...

Tomorrow 30 seconds and an ounce of conviction.

[Ben @ play @BenjaminSc7519]

@adocomplete Hey there Ado! I have a bunch of projects underway, that are being upgraded for 4.7 (image processing upgrades woot!!) and others net new - I guess, what I'm asking is, do you accept work in flight?

Or is the hackathon about net-new ideas?

[Ben @ play @BenjaminSc7519]

Opus 4.7 put in some serious work on https://t.co/CyW6XpZVwU over the past 48 hours, 2.5M tokens well spent

[Ben @ play @BenjaminSc7519]

@DanielleFong

[Ben @ play @BenjaminSc7519]

@felixrieseberg Felix, you had me at native api support. Claude, were rolling to microcenter!!