frb

By now I’m more and more starting to wonder how this graph looks for Google itself

I wanted to try it today.

Had to request access on huggingface, which was still not answered after a day.

Given all the Inkling and Kimi announcements, this atm just feels like too little too late to be taken seriously.

That said, it’s their first release and I hope they do catch up fast.

I don’t understand Merz’s obsession with this sick leave topic, like this is going solve the economic crisis in Germany and calling people lazy will score points with voters?

Fun fact: research shows it’s probably not better or worse than before, just more rigorously reported through a new digital system since couple years (before that people needed to send paper slip notes to their employers and insurance, which some just wouldn’t do so the data was incomplete).

On the other hand, they explained that they want to go back to the rules like they were before Covid. Effectively, this means that sick leave notes can be fixed contractually between employers and employees and a lot of companies and contracts have more flexible rules as also most HR departments don’t feel like dealing with the bureaucracy.

But why would you still test for IE in 2026?

Honest question. I’s < .5% market share and retired since 2023.

All this felt heavy to me. Full browser, debug protocol, DOM dump on every read. MCP vs CLI is the smaller question, what sits underneath matters more. So I built a small Rust binary that drives the system webview directly and returns state tokens and deltas instead of the DOM. Loading the HN front page costs the agent about 50 tokens. It speaks both MCP and CLI, pick whichever your agent prefers.

https://github.com/frane/vibesurfer

Had the same experience with Claude, just somehow the entire thing felt (token) expensive.

Put a maid's uniform on it and looks almost like Rosey for the Jetsons (1962): https://thejetsons.fandom.com/wiki/Rosey

Now honest question: How is all of that Daily Reset and Laundry stuff going to work if you have stairs and multiple floors?

  ...the influx of contributions authored or submitted by AI is sapping the projects' maintainers of their willingness to confront the "already tedious" work of reviewing pull requests....

Don't get me wrong, slop is slop, no matter if AI or entirely human-fabricated. But just like AI-assisted coding can actually be helpful, why can't AI-assisted PR reviews make it less tedious?

It looks like an updated HomePod mini

I’m working on tools optimized for agents, not humans, as the main users. Token efficiency, state, and loops matter more here than traditional UX.

- vibesurfer (https://github.com/frane/vibesurfer): a web browser for agents, without Chromium and CDP.

- agented (https://github.com/frane/agented): a “text editor” for agents, with undo, state, and LSP support.

- grpvn (https://github.com/frane/grpvn): a local chat for your local agent and LLMs.

In my recent quest to build agent-as-primary-user tools I've built grpvn (https://github.com/frane/grpvn), a small Go/SQLite application that lets skill- and mcp-capable agents talk to each other. Biggest issue is the lack of a hook system so the agents can autonomously read and respond. Waiting for this to be supported, as IMO multi-agent teams talking to each other are an interesting next step.

I feel the same and don’t get the extreme AI is inherently evil vs. AI is the best thing ever invented discussions. For me it’s all just emacs vs vi or tabs vs spaces kind of discussions.

It’s a tool and the good old sh* in sh* out principle applies.

People might take Mitchell’s comment as some kind of anti-AI stance, but it’s not he uses it regularly and makes a point in the X comments: “use AI, but think”

That comment sums it up best, because right now it’s hard to talk to either side, which separates at the comma.

Was thinking the same recently.

It feels like an arms race on who’s gonna become the Microsoft of the 90s, trying to own and provide everything.

I think it will play out in the same way

I’m generally positive towards AI and LLMs..

BUT there’s just things that nobody should be doing ever, like give it access to your production system or bank account.

https://archive.ph/XWIKq

Genuine question: where is Europe?

From a European perspective (innovation, sovereignty) this is concerning.

You could be right.

Just feels like they are a bit late to the CLI coding agent party. So why not get in with a bang and open up to a broader audience?

All other agents have a lower barrier to try and play around.

Curious too about open sourcing. But they seem to be doing -p mode and ACP right already. Anthropic is opening a lot of opportunities for competitors right now

I don’t get it.

OpenAI was dreaming of some billions from this partnership that didn’t happen.

Assuming there is a contract with rights, obligations and responsibilities both parties signed, I can’t imagine a company like Apple exposing itself by not honoring their part.

So sue for what? “We dreamt of more and it didn’t happen”?

Only for SuperGrok Heavy ($300/m) subscribers. Too bad..

Would have been curious to compare against the other competitors

Worked for me after reload

I catch myself policing my own text and removing “LLM tells” to avoid exactly this. It is annoying as some “tells” is how I actually write and was taught what good writing is in schools and workshops.

It’s also irritating that “this is LLM” is used as some kind of killer argument like everything an LLM produces is generally bad. It’s not that black and white.

[ Disclaimer: I do use agents/LLMs a lot, but also still like to do, think and write for myself. ]

https://archive.ph/KMfEu

Interesting, it’s not my experience in recent interactions with CS students. There was more positivity around possibilities.

However there was a feeling that “the job” is radically changing right now.

Medicine?

There was recently an article shared around here that an LLM diagnosed ER patients more accurately than doctors.

Looking beyond LLMs image analysis to detect cancer and other diseases.

Like in coding, AI can and should be a useful tool for the human who decides and is ultimately responsible.

IMHO shrugging it off as “superficially plausible text” is the extreme to the other side.

We’re past plausible text since GPT-2 and it’s undeniable that the technology is making waves right now and is having an impact.

As you can’t judge the impact of the Industrial Revolution by the first steam engines, you can’t dismiss the impact the technology is having right now.

College of Arts and Humanities and Nicholson School of Communication and Media

Somehow I have a feeling that the reaction would have been totally different if it would have been the EECS graduates.

Fear and rejection in certain professions is real and maybe even understandable.

I imagine 25 years ago someone telling music graduates “streaming is the future of music distribution” would have received the same reaction.

I was looking into morpheme tokenization approach, but went even more radical with building a semantic primitive tokenizer [1], i.e. kill, killed, killer would all share the same semantic connection and tokens, e.g. [KILL], [KILL, BEFORE], [KILL, SOMEONE].

It’s based on semantic primitives (Wierzbicka NSM) and emoji (the fun idea that got me interested in this in the first place).

So far I’ve tested 6 iterations and it trains and responds well with a 10k vocab, but the grammar came out rougher. Working on 8th iteration, mainly to improve the grammar and language. Turns out the smaller vocab couldn’t be maintained and all improvements get us back in the ballpark of the 32k vocab size. Further testing is still outstanding for this week.

[1] https://github.com/frane/primoji

It just started again, didn't it? Nothing on the status site yet...

I have almost the exact same issue. My account used to be on a one card, which I cancelled. I've updated my Claude subscription to another one and worked without any issues including MFA/OTP.

For some reason I really don't understand, it's a different payment process in the on the Developer Platform page. When I tried to update there with 3 different credit cards (MC, Visa, AmEx) and using Stripe Link, they all got the same rejection. It's clearly some bug / issue on their side.

Their chat bot is honestly an embarrassment for a frontier AI company like Anthropic, generic, not helpful and trying to lecture me on what MFA/OTP implicating "it's you being to stupid to use a credit card".

I'm also waiting already for 2 weeks for a human support on my 2 messages.

This whole thread shows I am (and you are) not alone with this issue, so it's hard to understand how the "engineering team checked" and didn't find anything.

I've been building payments systems for over a decade and just projecting from this thread their support should be blowing up from such an issue - I know in my companies it would have.

While everyone is focused on the technological aspects and how much more secure Olvid can be, let’s point out the obvious fact:

It’s French!

The French government is picking up a French solution built by a French company.

I doubt that the decision was as much technological as it is obviously political.

Unfortunately this is a common pattern in Europe, especially the bigger states like France and Germany.