Danil

Software developer, living and working :-)

[Modern Software Engineering @ModernSoftwareX]

Anybody else ever get this feeling? 😬

[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.

[Kent Beck 🌻 @KentBeck]

Dario demonstrating that he doesn't understand software engineering. The human side of what we do has always been the heart. I can empathize with *wanting* to never talk to an engineer ever again, but engineering becomes more important with better tools.

[Siva @sivalabs]

Looks like AI honeymoon period is over 😉

Now everyone is slowly putting token limits and switching to API usage based billing.

#FreeLunchIsOver

[Sonar @SonarSource]

📋 GPT-5.5 generates a lot of code, and that code is complex with minimal comments. The faster a model generates, the more a team has to verify. That gap is real, and it's where issues land.

[Tesla Australia & New Zealand @TeslaAUNZ]

FSD Supervised testing in Auckland, New Zealand

Judder bars and all

[Danil @dmatsynin]

@old_sound Not New Zealand :)

[Danil @dmatsynin]

@wrumsby Oh yes! I did some reading on the Crew/Cockpit Resource Management some time ago and some concepts could be applied to the incident management in IT almost like for like.

[Danil @dmatsynin]

@wrumsby Not strictly a postmortem related though, so maybe it’s not that.

[Charity Majors @mipsytipsy]

There are behaviors and conditions, user experience and flows. Bugs are just extreme edge cases. There is sooooo much more to understand.

You can't make a copy of production. Only prod is prod. But this doesn't mean you have to ship dangerously, or expose users to scary shit.

[Danil @dmatsynin]

@AklBikes Oh wow

[Danil @dmatsynin]

@ChrisKeall Practicing “tactical approach “? 😀

[Danil @dmatsynin]

@alexpotato Wouldn’t this be an example of the “Manage your manager” wisdom?

[Danil @dmatsynin]

@talios Seems to be working ok. A momentary blip perhaps?

[Danil @dmatsynin]

@parsley72 Can’t argue with that 😀

[Danil @dmatsynin]

@wrumsby Double cab 4 liter diesel Ute that is, nothing else permitted.

[Charity Majors @mipsytipsy]

We talk a lot about modern best practices for software engineering. In my opinion, the big ones are:

* devs owning their code in prod
* continuous delivery in a tight loop
* using feature flags to separate deploys/releases
* o11y-driven development
* testing in production

[Charity Majors @mipsytipsy]

The hardest part of software has ALWAYS been running it, maintaining it, and understanding it -- not writing it. But this was obscured for a long time by the upfront difficulty of writing code.

Generative AI turns this upside down, and makes it clear where the real costs lie.