nialse

4 ms·

31.
by nialse 5mo ago
”Religious suffering is, at one and the same time, the expression of real suffering and a protest against real suffering. Religion is the sigh of the oppressed creature, the heart of a heartless world, and the soul of soulless conditions. I
32.
by nialse 5mo ago
Keen observation. Maybe automation will come for the AI as well?
33.
by nialse 5mo ago
The people were longing for utopia, just not the same utopia.
34.
by nialse 5mo ago
Another argument for less human-like AI then, I guess.
35.
by nialse 5mo ago
I'm starting to long for the age after AI. When the generative euphoria has settled and all outputs are formally verified based on exquisite architectures and standards.
36.
Thoughts on Doctoral Student Recruitment(youtube.com)
2 points by nialse 6mo ago|0 comments
37.
Automated systematic literature review with Claude Code(youtube.com)
2 points by nialse 6mo ago|0 comments
38.
by nialse 6mo ago
I see your point. Many of my prompts for reasoning ends with: No code. Planning mode is sort of the workaround for this specific situation. Sometimes it is useful for the AI agent just to think. It looks like I need a screwdriver in additio
39.
by nialse 6mo ago
To be fair, there is likely not much training data on the difficult conversations you need to handle in a senior position, pushback being one of them. The trouble for the agents is that it is post hoc, to explain themselves, rationalising r
40.
by nialse 6mo ago
I think that it is a fair perspective to allow role play, and it's useful too, when explicit. Does not really make sense for AI to cosplay human all the time though.
41.
by nialse 6mo ago
Maybe we should change that? Of course symbolic AI was the holy grail until statistical AI came in and swept the floor. Maybe something else though.
42.
by nialse 6mo ago
The other day Codex on Mac gained the ability to control the UI. Will it close itself if instructed though? Maybe test that and make a benchmark. Closebench.
43.
by nialse 6mo ago
Asking for code to manipulate the AST is another route. In python it can do absolute magic.
44.
by nialse 6mo ago
Interesting. When I code, I want a boring tool that just does the work. A hammer. I think we agree on that the tool should complete the assignment reliably, without skipping parts or turning an entirely implementable task into a discussion
45.
by nialse 6mo ago
That is probably the next step, and in practice it is much of what sub-agents already provide: a kind of tabula rasa. Context is not always an advantage. Sometimes it becomes the problem. In long editing sessions with multiple iterations, t
46.
by nialse 6mo ago
Agreed. I am asking for something beyond the current state of the art. My guess is that stronger RL on the model side, together with better harness support, will eventually make it possible. However, it's the part about framing the fa
47.
by nialse 6mo ago
It's not necessarily "ignoring" instructions, it's the ironic effect of mentioning something not to focus on, which produces focus on said thing. The classic version is: "For the next minute, try not to think about
48.
by nialse 6mo ago
Agreed. We should not be anthropomorphising LLMs or having them mimic humans.
49.
by nialse 6mo ago
Codex in this case. I didn't even think about mentioning it. I'll update the post if it's actually relevant. Which I guess it is. EDIT: It's specifically GPT-5.4 High in the Codex harness.
50.
by nialse 6mo ago
On point. I'm more interested in what comes after LLMs/AI/AI-agents, what the next leap is.
51.
Less human AI agents, please(nial.se)
161 points by nialse 6mo ago|170 comments
52.
by nialse 6mo ago
My point was narrower than suggested. If Mythos is in fact a security risk, then the NSA is one of the actors most likely to already understand that. The surprising part is not that they would evaluate or use it anyway, but that we are hear
53.
by nialse 6mo ago
That is expected. What is not expected is us knowing about it. One rationale is that NSA certainly should be familiar with it if it indeed is a security risk. Nothing to see here.
54.
by nialse 6mo ago
This pushed me to try pi.dev with a remote ollama instance running gemma4:26b on a 24GB RTX 3090. It is not perfect, but it's quite okay-ish with 128k context length. It's not a model primarily for coding I guess, but IIWIW. Thank
55.
by nialse 6mo ago
Manufacturing demand. Create the problem. Solve the problem. The 2x allowance last month was not of the goodness of their hearts. It’s pumping up the demand AND the revenue simultaneously. The squeeze before an IPO. How many of us have look
56.
by nialse 6mo ago
- Yes it is! - No, it isn’t. Is it intelligence to attribute this quote to Monty Python, or isn’t it? Remembering things is traditionally considered a part of intelligence, memory that is. The subject has been discussed for ages. Personally
57.
by nialse 6mo ago
In terms of communication: The only clearly communicated message is that TDF is not fit for fulfilling its purpose and likely never have been. As an outsider I would suggest ceding the project and IP to a third party not involved in the his
58.
by nialse 7mo ago
Paranoia. And also ironic considering their base LLM is a distillation of the web and books etc etc.
59.
by nialse 7mo ago
Going forward this may become somewhat the norm. The main thing that has stopped major rewrites or reimplementations under other licenses is the amount of cognitive resources that is needed. Many companies have made similar efforts just to
60.
by nialse 7mo ago
Microsoft injecting permanent ads in PRs? Has this been independently confirmed? Brought to you by Carl’s Jr.

More ›