jbergqvist

6 ms·

1.
by jbergqvist 5mo ago
Doesn't this whole argument fall apart if we consider iteration over time? Sure, the initial implementation might be uncoordinated, but once the subagents have implemented it, what stops the main agent from reviewing the code and sorti
2.
by jbergqvist 5mo ago
"Helped build itself" is a bit of a stretch here, it makes it sound as if the model was doing lasting self-improvements. What the article describes is that the model was able to tweak to its own deployment harness (memory, skills,
3.
by jbergqvist 5mo ago
In my experience, one reason for unnecessarily complex solutions during vibe coding is the incremental work pattern. Most users don't spend much time designing the solution, but instead jump quickly to implementation and then iterate.
4.
by jbergqvist 5mo ago
When I want to solve a new problem with an agent, I always ask it to search broadly for prior work in the given area online, and then analyze if we can build our solution using it as inspiration. I see it as the solution being out there in
5.
by jbergqvist 5mo ago
Limit access to whatever their project requires. The difference is that human interns have some common sense and won't suddenly be hijacked by a hidden message they stumble upon while searching the web, instructing them to exfiltrate a
6.
by jbergqvist 5mo ago
This works well for vibecoding on a codebase in isolation, which to be fair is what the author is addressing. I don’t think it solves the problems at the current frontier of agent use though, where you expose internal infrastructure via too
7.
by jbergqvist 5mo ago
Maybe. Personally I find it hard to tell how sincere this is. The cynical take is that this is just an attempt to secure their own position, especially if AI progress slows down and competition increases. However, if it does not and we’re t
8.
OpenAI: Industrial Policy for the Intelligence Age(openai.com)
8 points by jbergqvist 5mo ago|2 comments
9.
by jbergqvist 5mo ago
Usage limits are more generous and GPT 5.4 is a good model, but yes, UI/UX lags behind Claude Code. Currently I'm especially missing /rewind with code restoration and proper support for plugin marketplaces
10.
by jbergqvist 6mo ago
NemoClaw is an OpenClaw security wrapper, not a replacement
11.
by jbergqvist 6mo ago
Not to give Snowflake credit for a design that clearly wasn't a sandbox, but I think it's worth recognizing that they probably added the escape hatch because users find agents with strict sandboxes too limited and eventually just
12.
by jbergqvist 6mo ago
I've found that spending most of my time on design before any code gets written makes the biggest difference. The way I think about it: the model has a probability distribution over all possible implementations, shaped by its training
13.
by jbergqvist 6mo ago
Does this matter in practice though? By modifying some of the generated code and not taking a solution produced by an LLM end-to-end but borrowing heavily from it, can't a human claim full ownership of the IP even though in reality the
14.
by jbergqvist 6mo ago
I think AGENTS.md will still have a place regardless. There are conventions, design philosophies, and project-specific constraints that can't be inferred from code alone, no matter how good the judgment
15.
by jbergqvist 6mo ago
Would have loved to see a more detailed breakdown of performance by task type. The commit metadata is right there, seems straightforward to tag commits as feature vs refactor vs bug fix vs API change and report per-category numbers.
16.
by jbergqvist 6mo ago
Producing the most plausible code is literally encoded into the cross entropy loss function and is fundamental to the pre-training. I suppose post training methods like RLVR are supposed to correct for this by optimizing correctness instead
17.
by jbergqvist 6mo ago
This would be my guess too. It can probably be generated synthetically or via agentic rollouts, but high quality long context examples where outputs meaningfully depend on long-range interactions probably remain scarce
18.
by jbergqvist 6mo ago
This seems like a win for open source maintainers pressed on time and resources. Whether or not LLMs find novel security risks or just pattern-match known issues, many vulnerabilities are discovered late (or never) simply because nobody has
19.
by jbergqvist 7mo ago
In a way, isn't this the same old data moat that always existed in AI/ML, but supercharged? Generalist models can now reason over proprietary data as context instead of requiring you to train narrow expert models on it. What chang
20.
by jbergqvist 7mo ago
Very interesting benchmark, excited to see what comes out of this. Considering humans are enourmously more sample efficient compared to today's models, it seems clear there's a lot of room to close that gap. The fact that they hit
21.
by jbergqvist 5y ago
One could also argue that if the developers did that, the market value of the digital item in question would drop. The value that the buyer receives is grounded in the large time investment required to acquire the item in the game. Even tho