novaRom

This reminds me what professor Geoffrey Hinton did talking many times in the past. Super intelligent artificial aliens.

I think the most interesting take is to look at their thinking logs to find out how exactly they decide what weight to change and why.

They blocked anonymous readings of old.reddit few months ago, rumors are they will do the same for main reddit. A kind of moat to keep control on who's reading what.

That's the point I am really convinced today that few small size agents actually better than one remote begemoth. I switched fully to local models (from claude, sol, and deepseek). A 27B model can be served on any modern PC or laptop with 64 GB or RAM (or VRAM on non-Macs). A bunch of Qwen coding agents is fantastic with opencode, works great also with pi. The only disadvantage is required compaction due to limited 256k context size, so it stops for few seconds once every hour or so. I am able to produce remarkable things similar to "see what I vibe-coded" posts here on HN but using only open weights, open source, and run everything locally.

It feels a lot of reddit comments (but also here on HN) are AI generated. It would be helpful to see a kind of marking of probability a comment is generated.

I have switched to Mistral free app just from US apps completely recently - it beats for me absolutely all alternatives:

ChatGPT - stopped using it like a year ago, it feels they deliberately made it ridiculously stupid (for fun?). I know, they were first, but today it's a joke.

Claude - tried it out of curiosity few times, but it's close to nothing when it's about non coding questions. I don't understand how anyone can like its gibberish answers?

Gemini - like many, switched to it like a year ago and it was best choice for daily usage. Nice, clear answers, never nonsense. Somehow they made it faster recently, but answers have become shorter, and it feels slightly stupid (but yet smarter than Claude or ChatGPT).

This was my motivation to try Mistral and I am absolutely amazed how good it is. It answers "in full", smart, search, research, full assessment, examples - looks very much like Pi/Opencode with subagents forking here and there (maybe OpenClaw is what they use?).

I am based in EU, so maybe others geographicly experience is different - again it' all no-subscription just free apps.

It looks specialized models of 27B to 52B active size are currently state of the art.

I think we will not see any significant increase of that numbers in next 4 - 5 years.

Does MoE depth justify itself to be stored and served as "one" model for particular tasks like agentic coding vs same active params non-MoE model?

For me, this result is significant. It basically adds much more hope P is equal NP (that majority seems being against). I believe one day AI will explain a major flaw in modern mathematics, so a kind of correction in axioms will be suggested to prove Rieman's hypothesis (or a completely new apparatus)

> everyone but the US is now looking to divest from them as fast as possible

This! EU is now in full "decoupling" mode, many future contracts are now under reviews and plans how to preserve sovereignty.

Btw seeing what ArtCraft does it is now really realistic. Will it be SuSe + Mistral, or a kind of ArtCraft like vibe coded solutions?

Can this company ever be trusted?

Claude Fable 5's model card disclosed safeguards that would invisibly degrade output for requests linked to frontier LLM development. The research community called it "sabotage". Dean Ball: "degrading performance on ML research without telling the user is shockingly hostile" — including because it would corrupt third-party safety evaluators of Anthropic's own models.

What is unique about them? What is smth they have nobody else can reproduce?

It looks here in EU most talks recently about souvereign AI and how to quit digital dependence on oversea providers, at least first attempts to decouple. Thus I wonder if is there a place for new serious competitors?

The AI tech itself is likely very liberating - it seems allow to reproduce anything digital with less efforts than ever before, so others may simply replicate your business quickly?

Anthropic has a large in-house PR machine (80 people). Main functions are "attention flooding", well-resourced news-blitz strategy.

Circular financing at times of high inflation and geopolitical instability is risky. Company A invests in Company B, B invests in C, and C invests in A. If A is forced to sell its shares in B to raise cash, B's price drops. Now B is worth less, so C now loses money too. Self-reinforcement.

Why? The answer is directly in this article, : "for women particularly, the idea of marriage might not seem so appealing, in a culture where they were expected to bear most of the responsibility for household chores"

Btw, is it good or bad for individuals and for society?

Some limited intelligence but strictly focused on the environment and goal. Similarly, local 27B models often outperform large 816B (42B active) models in different tasks.

> Microsoft has completely lost the ability to make good software

did they ever? maybe MS DOS, Win 3.1, and Win2000 were ok, although PC DOS and OS/2 were better; Sure what an unusable junk it's today is beyond any precedent in history, but why many companies use it today at all? why?

If the threat frontier companies describe is real, transparency is the cheapest insurance society can buy. Let them open their training data, I bet it's a lot of questions.

They are very nervous about "no moat anymore", so each week like those hysteria making news in order to establish self-regulated duopoly. What next proposal will be to ban open models? What a joke society has become, delegating its future into hands of monopoly seeking felons (training on copyrighted stuff, training on users' chats - see Naive-Stoke discovery).

I was opencode user, but like few months ago tried pi and never looked back. Pi with open models is absolutely mind-blowing - my agents run with speeds not achievable with any remote frontier model - they certainly not as smart, but it doesn't matter because they do their "work done" perfectly and fast.

All problems I had with opencode are not presented in pi - it looks it's the best approach to guarantee most productive coding environment - speed, simplicity, transparency, absolute customization. I will certainly try this 2.0, maybe they have now improvements.

Why is it not mandatory yet to disclose full trainig data used to produce any new model? Without it we cannot evaluate any product's safety.

They may talk about rogue agents, sandbox escapes, but this will not solve the core issue: secret training recipes are a security hazard in themselves. Full transparency is needed.

Nothing new, I always see attempts to cat ~/.ssh/*

Of course agents work in a strict sandbox, they always seem to understand it quickly on quite first turn; but they continue like trying to access something belonging to the host system despite they know they are in a sandbox.

Frankly many people run coding agents absolutely without any containment, entire companies - all their keys, code, data, trade secrets, everything goes to remote LLMs.

I cannot understand how is it legal to allow AI companies serving remote LLMs to steal users' data? Frontier models train on users' data, right?

Upvote. Also ban collecting, storing, and using users' data for training models (see how Navier-Stoke millenium problem solution was produced). But not just data, also recipes how this data is used to train models need to be released (see how HF was hacked).

If we really concerned about data privacy and concentration of power then it's logical to ban closed models instead.

It's just circular financing - we give you GPUs, you give us your shares of a promised future $ 2,000,000,000,000 valuation. Despite they are unprofitable today and tokens are getting cheaper and cheaper.

why "after" and not "before"?