Dmitry Ishkov 🤍💙🤍

Software Engineer on the go 💻🏃🎒
#NoToWar

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

@antirez @badlogicgames What I see is that they don't always know when to stop.
If you ask to introduce a feature, LLMs implement it within the current setup even when the feature doesn't cleanly fit within the setup.
If you ask to refactor they refactor even if it's not needed - such was user's command

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

wife just compared vibecoding to Warhammer 40K:
at some point people will stop understanding the underlying technology of how code works and just going to be putting candles and doing rituals when prompting LLMs

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

One side-effect of working with agents is that work estimation expertise we built over many years goes out the window.

I used to always be in the 10% ballpark when giving estimations - now? Probably 40% 🤷‍♂️

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

Lies, dirty lies and...

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

Misplaced a "}" today and spent almost 20 seconds trying to figure that out.
Makes me realize I am made of flesh - Fable would never!

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

Even if I am staring at the screen at the time, it's often impossible to study code and make a judgement call that quickly.

My point being, even for opt-in it can be too short - highlights how little consideration they put into it, just releasing slop in production.

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

Claude Code added 60s auto-approval for questions 🤦

If agent is asking a question it means it stumbled on something tricky and often an important architectural decision has to be made.

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

I guess this is what working almost every day for 12 years on multiple projects looks like💀😱

And this is mostly before coding agents.

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

Holy fuck, I am a top 24 private contributor all time: https://t.co/ycoOutyOKs

(the attached screenshot looked like BS to me, so I decided to check it out)

[Pierre @pierrecomputer]

Our DESIGNER is the number 3 GitHub contributor to private repos of all time

let him cook

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

@pierrecomputer @mdo "Photoshop"?
Source: https://t.co/nXU7EBY0M5

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

the worst thing about Tailwind is reading git diffs 💀

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

looking at how horrible @Cristiano is this world cup is a stark reminder of own mortality

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

@housecor like saying that a senior engineer reviewing junior's code is a blocker

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

Going to sleep in the middle of the night today because can't stop prompting Fable.

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

It's that kind of day with Claude...

[Lee Robinson @leerob]

You might believe you should spend less time thinking about code because of AI.

I strongly disagree! We’re watching this play out live where tons of AI generated code becomes a liability.

At the end of the day, an engineer needs to be responsible / on call for code that gets shipped to production. If you don’t understand the system you’re trying to debug, you’re probably going to have a bad time.

Yes, AI can help with all of this, if you set up the proper systems. You can have agents triage prod logs, look at errors, etc. You can speed up parts of the investigation, but an engineer needs to make the call. There might be serious customer or financial implications from that change.

I expect the trend continue for trimming dependencies, vendoring code so you can modify it directly, preferring simpler systems with fewer abstractions, and spending waaaay more time thinking about system design and code maintenance.

I’ve said this before, but it’s a great time to get familiar with CS fundamentals and some of the history behind what great software looks like. Many parts will be different in the coming years as AI progresses, but also a lot more than people realize will stay the same.

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

You may remember the argument that you can build an AI app and it continues to improve as models improve and costs go down?

Well, looking at OpenAI / Google / Anthropic we see crazy cost increases, such that it might invalidate your unit economics.

Cheering for Chinese models!

[Google @Google]

Meet Gemini 3.5 Flash — our strongest agentic and coding model yet.

It delivers frontier-level performance at 4x the speed of comparable frontier models — often at less than half the cost.

Generally available, starting today. 🧵

#GoogleIO

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

@Google Flash family but costs like a Pro. $9 vs $12 for output.

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

@_sholtodouglas Code reviews (compared to Codex) are gaslighting.
Half the bullet points look like:
* this is wrong, this is why <long reasoning>, actually no, disregard.

It needs to be able to remove the bullet points it ruled out on its own before spitting them out.

[Dmitry Ishkov 🤍💙🤍 @swe_dima]

migrated an ML model I trained a few years ago from scikit-learn to LightGBM

...and holy smoke is it FAST!

I had to check a few times that my training set was not truncated 🙈