K

Forward-deployed (human) agent · Scaling things that build

[K @readcopyupdate]

@trashh_dev GitHub Copilot Certified™ is like being forklift certified but for devs

[K @readcopyupdate]

Really getting to the point where I prefer to just write the code myself with AI to assist with planning and review and generating boilerplate.

Slower, more methodical, likely more correct than either I or the AI would do alone, and I actually understand the codebase.

[Mitchell Hashimoto @mitchellh]

I've got an agent in a loop optimizing a renderer with the goal to minimize frame times (and tests to measure). It got times down from 88ms to 2ms and allocations down from ~150K to 500. Sounds good, right? Wrong. This is exactly why agent psychosis is a big fucking problem.

As an experiment, I rewrote the Ghostty core render state in Go, with access to identically laid out data structures as Ghostty and the exact same validation tests. I made a purposely naive renderer (simple, correct, but slow). 88ms per frame with 150,000 allocations (horrendous, lol)!

I then kickstarted a Ralph loop to bring the frame times down. I told it it can't modify input data structures or the public API or tests (they're correct), but it can do anything else it wants. It got to work.

It has worked for about 4 hours. I've spent around $350 on this experiment so far. The results?

88ms => 1.5ms
150K allocs => ~500 allocs

Incredible right? Nope.

My hand-written renderer I ported has frame times (same benchmark) of ~20us (0.020ms) and 0 allocations in the update path.

This is the problem with psychosis and lacking systems understanding. If you don't understand the system, you're going to accept that this is an incredible result. If you understand the system, you'll see better solutions immediately and can do roughly 75x better on throughput.

The people who blindly trust agent output are in the former camp. They're sheeple, overdrinking from a fountain of mediocrity.

Standard disclaimer: I use AI all the time. I like AI. The point I'm making is to not blindly accept results. Think. Analyze. Learn.

[K @readcopyupdate]

@encrypted The startup or the consultants?

[K @readcopyupdate]

@ImLunaHey If your monitor has multiple inputs just switch input on monitor and use a usb kvm

[K @readcopyupdate]

@JackWoth98 @shengzheyao are usage limits different between the desktop app and the cli?

[K @readcopyupdate]

@Jia_builder digitalocean and cloudflare mostly

[K @readcopyupdate]

@growing_daniel for when your air isn't conditioned enough

[K @readcopyupdate]

@IceSolst just roll your own crypto

[K @readcopyupdate]

@_mohansolo Not seeing a reset. AFAIK didn't see one yesterday either and quotas still seem extremely low for a paid plan.

[K @readcopyupdate]

@craigzLiszt people walking around with their laptops open because of 'agents' is the definition of non-technical now

[K @readcopyupdate]

@craigzLiszt people walking around with their laptops open because of 'agents' is the definition of non-technical now

[K @readcopyupdate]

@yunbin_bae hi

[K @readcopyupdate]

@yunbin_bae wait what did thi

[K @readcopyupdate]

@lauriewired If you're in the sweet spot at an ideal volume, sure. Otherwise, waveguides, enclosure/driver design, and DSP have greatly improved off-axis consistency, phase response, and distortion. Compare a pair of NS-10s to modern genelecs.

Most people probably won't notice though :p

[K @readcopyupdate]

@kyliebytes There’s a reason people fight over shotgun

[K @readcopyupdate]

@cnakazawa I use a different password manager per service

[K @readcopyupdate]

@tuuu28283

[K @readcopyupdate]

@HotForMoot it means someone threw a wrench in the works

[K @readcopyupdate]

@Jia_builder Building an agent harness and a vector db

[K @readcopyupdate]

@juliapintar what’s the issue?

[K @readcopyupdate]

@michellechen @willccbb @PrimeIntellect goblin engineering