Dominik Seifert

⏰🚀 Time-travel debuggerer @replayio
🤖 Working on AI x Program Analysis / Code Comprehension / Debugging
🎓 CS PhD

[Dominik Seifert @DominikSeifert4]

Claude doesn't understand rg.
It used `rg -rn X ...` which replaces all of "X" with "n".
Then it tried to explain how it understands everything, while literally chasing ghosts.

It should probably go and learn from composer.

[Dominik Seifert @DominikSeifert4]

It must be 1999 again, or why else would Cursor, Claude-code et al bring back f*king popups?

[Dominik Seifert @DominikSeifert4]

Upset about GitHub's recent enshitification?
Cursor has been acting up in diverse ways 3 days in a row (different each day), to the point of unusability...

[Dominik Seifert @DominikSeifert4]

I wrote a harness that introduced its own DSL with algorithmic agent definitions as first-class citizens.

Language is interpreted by an agent. Works but waiting 2 minutes for a few lines of code is a bit much.

Time to try out @resonatehqio (cc @AndaristRake, its time!)

[Dominik Seifert @DominikSeifert4]

Claude is disappointingly bad at GIT-FU.

Problem: Find the latest upstream rebase SHA for our runtime forks.

Results:
* ChatGPT comes up with the correct solution and fast. Proposes simple programs to do it.
* Claude doesn't get there at all.

[Dominik Seifert @DominikSeifert4]

Anyone notice an uptick in usage of the words "pin" and "envelope" with Opus 4.7? ✉️📌

[Dominik Seifert @DominikSeifert4]

Of course swe-bench was part of Claude's training. If that is not a significant indicator of how meaningless swe-bench is these days I don't know what is🤓

[Dominik Seifert @DominikSeifert4]

Truely a great time to be alive. I can't believe that I don't have to debug this myself anymore...

[Dominik Seifert @DominikSeifert4]

Schroedinger's Claude: Understanding everything, but also absolutely not, at the same time.

> "OK. Now I'm clear on the full chain. Let me trace the exact paths:"

[Dominik Seifert @DominikSeifert4]

2026 prediction:

Agentic coders will need to become experts in problem decomposition.

Those who try to solve everything in a single prompt will not thrive.

[Dominik Seifert @DominikSeifert4]

Hot take: `Extended thinking` pollutes the context window.

Example:
* Ask Claude Opus 4.5 to generate some specific code.
* It generates a `thinking` event with *all* the code.
* Then it generates an exact duplicate of that very same code in a separate `thought` event.

[Dominik Seifert @DominikSeifert4]

Yesterday, I asked @AndaristRake if he has a suggestion as to how to raise a TypeScript error if some function does not enumerate all relevant combinations of a union.
Today, I wake up to a PR for my code and a PR for the typescript compiler (https://t.co/n5DONxipE6) 😁👏

[Dominik Seifert @DominikSeifert4]

@NotionHQ I think I understand now why Notion has stopped working entirely for me for almost a week now

[Dominik Seifert @DominikSeifert4]

@XiangruTang Love to see it. Fault/code/feature localization is one of the big unsolved problems and its not getting enough attention. Also agreeing with @acrognali about the benchmark. But, are you going to share `Loc-Bench`? I did not see a link/reference in the paper.

[Dominik Seifert @DominikSeifert4]

📣 News: 3 new *big* `Code Agent` competitors appeared just this week, yet none of them can accurately tell me how my code executed 🤔

* Claude Code: https://t.co/dICOANAwey
* Copilot Agents: https://t.co/cl3hxVdifv
* Gemini Code Assist: https://t.co/MoLcMcCIsZ

[Dominik Seifert @DominikSeifert4]

@_marco @OpenAI You seem to agree with them then. According to the paper "Most tasks [...] are bug fixes" 🤷‍♀️

https://t.co/EBw2iGd2Fh

[Dominik Seifert @DominikSeifert4]

Trying to learn about GPRO vs. critic model-based (e.g. PPO) policy optimization. I asked both 4o and DeepSeek (online version).

1. DeepSeek started out with more concise explanations.
2. But when asking for examples, DeepSeek chokes, while 4o provides examples no problem.

[Dominik Seifert @DominikSeifert4]

That's a long list 👀

[Dominik Seifert @DominikSeifert4]

@kevinkern 1. Its pretty cool that Gemini already has planning built into their web offering. Thanks for sharing. Gpt and everyone else will probably/hopefully offer that as an option as well.
2. Note that 4o has tools and web access. o1 is supposed to also get that, soon™

[Dominik Seifert @DominikSeifert4]

Prediction: In 2025, coding agents and assistants like OpenHands (@allhands_ai), Devin (@cognition_labs), @cursor_ai (and many more) will start using traditional debuggers. Not sure how well that will go🤔