IMTDb

I’m curious what’s behind the speed improvements. It seems unlikely it’s just prioritization, so what else is changing? Is it new hardware (à la Groq or Cerebras)? That seems plausible, especially since it isn’t available on some cloud providers.

Also wondering whether we’ll soon see separate “speed” vs “cleverness” pricing on other LLM providers too.

The idea is that, over time, the quality and accuracy of world-model outputs will improve. That, in turn, lets autonomous driving systems train on a large amount of “realistic enough” synthetic data.

For example, we know from experience that Waymo is currently good enough to drive in San Francisco. We don’t yet trust it in more complex environments like dense European cities or Southeast Asian “hell roads.” Running the stack against world models can give a big head start in understanding what works, and which situations are harder, without putting any humans in harm’s way.

We don’t need perfect accuracy from the world model to get real value. And, as usual, the more we use and validate these models, the more we can improve them; creating a virtuous cycle.

A significant portion of engineering time is now spent ensuring that yes, the LLM does know about all of that. This context can be surfaced through skills, MCP, connectors, RAG over your tools, etc. Companies are also starting to reshape their entire processes to ensure this information can be properly and accurately surfaced. Most are still far from completing that transformation, but progress tends to happen slowly, then all at once.

He intentionally disrupted all communications including, but not limited to, emergency ones. In process he tried to unilaterally control a public resource based on his own authoritarian view of what is "good" vs "bad". He is lucky he is not in jail.

Maybe a dumb question on my side; but if you are using a GUI like emdash with Claude Code, are you getting the full claude code harness under the hood or are you "just" leveraging the model ?

> The initial software Pump and Dump event could be considered when Cursor burned through millions of dollars to build a barely working browser. Naturally there was no way to finish such a monstrous heap of software into a working product and why would anybody use a vibe coded browser anyway? The "dump" on their end was to use this as marketing bait and a way to inflate their valuation.

Let me introduce you to the wonderful world of "research." It's what happens when you're willing to spend money on things without immediate, obvious ROI. The real value often comes not from the resulting product, but from the lessons learned along the way. I also don't see what's wrong with showcasing the results of your experiments. How many developers have implemented a toy ray tracer and put it on their personal GitHub? No one in their right mind believes Pixar will use it for their next renderer, but should we conclude those people are inflating their CVs with bait? Or can we acknowledge it's a cool project to undertake, and pulling it off requires real skill? If individuals are welcome to do this, why can't organizations? I want to see more "we did a fun thing, here are the results." There's a playfulness in that approach I find refreshing. Just because it comes from a for-profit company doesn't make it cynical.

If it's that easy to not need the banners, I'd expect EU websites themselves to lead the "no cookies needed" movement.

Yet https://european-union.europa.eu displays a cookie banner for tracking on what is essentially a static informational site. If the EU itself feels tracking is valuable enough to warrant the banner on their own pages, it's hard to fault businesses (whose survival actually depends on understanding their audience) for making the same choice.

At least they're compliant with their own regulation, I suppose.

A key difference is that the cost to execute a cab ride largely stayed the same. Gas to get you from point A to point B is ~$5, and there's a floor on what you can pay the driver. If your ride costs $8 today, you know that's unsustainable; it'll eventually climb to $10 or $12.

But inference costs are dropping dramatically over time, and that trend shows no signs of slowing. So even if a task costs $8 today thanks to VC subsidies, I can be reasonably confident that the same task will cost $8 or less without subsidies in the not-too-distant future.

Of course, by then we'll have much more capable models. So if you want SOTA, you might see the jump to $10-12. But that's a different value proposition entirely: you're getting significantly more for your money, not just paying more for the same thing.

It is not in the pursuit of happiness that we find fulfillment, it is in the happiness of pursuit - Denis Waitley

Off base.

The only way to get "perfect rating" is to go to your junior dev and bring another interruption (maybe the dev was 90% done !). So now he has been interrupted twice by two different manager and you have contradicted your own boss in front of an employee. You just broke a cardinal rule of middle management: it's ok to tell your boss he is wrong, but not in front of someone else. Additionally, you also need to tell him to f** off with is request to get the numbers (without even trying to understand if the request was legitimate or not !), so that your precious sprint is saved. I don't see how he gets what he wants in your ideal handling. AT best you seem to tell him you will "look into it" in two weeks.

Much better solution is to help you junior dev solve the problem so the interruption goes away as fast as possible and he can go back to contributing to the sprint. If the VP requires these numbers and went as far as back channeling you there is probably a quite good reason for that. Maybe the last time he needed something you told him it was not possible because the sprint thingy is unmovable ? Once you have the result, you can go give those to the VP yourself, highlight the work of the junior dev, and use this "I am giving you the very important data you asked for" as a foot in the door to show that you had to pull the dev from the other feature, that these interruption also have cost and that you are more than happy to take care of them. He gets what he wants, the difficult conversation of "you did not do what you are supposed to do" happens behind closed doors, and you have a much better of getting results if he sees you as an ally to get his important stuff rather than a hinderance.

Companies can become « too big to fail » and dictatord can become « too powerful to fall » ?

You are comparing apple to oranges. Social media posts are static, don’t watch you etc. But the distribution platform does all these things.

In books it’s exactly the same thing: do not believe for one second that the publishing industry does not watch engagement metrics (aka: sales) and does not adapt to the taste of the market. It’s also tuned to maximize outrage; see how popular unauthorized biographies of polarizing figures have become - who is next on Walter Isaacson list ? I am betting Trump must be somewhere there and it’s gonna be a banger.

If these customers have set an automated system to pay you $200 every single minute, that’s correct. If they haven’t and it was just a one off sale, you are missing the “recurring” part in ARR.

> It seems like EU in general should be heavily invested in Mistral's development, but it doesn't seem like they are

The EU is extremely invested in Mistral's development: half of the effort is finding ways to tax them (hello Zucman tax), the other half is wondering how to regulate them (hello AI act)

> Our first and default tool should be some form of lightweight automated testing

Manual verification isn't about skipping tests, it's about validating what to test in the first place.

You need to see the code work before you know what "working" even means. Does the screen render correctly? Does the API return sensible data? Does the flow make sense to users? Automated tests can only check what you tell them to check. If you haven't verified the behavior yourself first, you're just encoding your assumptions into test cases.

I'd take "no tests, but I verified it works end-to-end" over "full test coverage, but never checked if it solves the actual problem" every time. The first developer is focused on outcomes. The second is checking boxes.

Tests are crucial: they preserve known-good behavior but you have to establish what "good" looks like first, and that requires human judgment. Automate the verification, not the discovery. So our first and default tool remains manual verification

I know few countries that reject poisonous US social media in favor of better platform that is safe for children, safe for news and information, and safe for society and for Democracy itself: the peoples democracy of North Korea, the democratic republic of Iran, the not authoritarian society of Russia, etc

I see tremendous correlation between restriction of access to some websites and straight up dictatorship that pretend to protect it's population from the evils of foreign influences.

Or maybe it’s just strange classification. I see a lot of prompts on the internet looking like “act as a senior xxx expert with over 15 years of industry experience and answer the following: [insert simple question]”

I hope those are not classified as “roleplaying” the “roleplay” here is just a trick to get better answer from the model, often in a professional setting that has nothing to do with creative writing of NSFW stuff

> or example, we can forbid corporations usage of algorithms beyond sorting by date of the post

Hacker News sorted by "new" is far less valuable to me than the default homepage which has a sorting algorithm that has a good balance between freshness and impact. Please don't break it.

> It is you that may have misinterpreted what regulations are.

The definition of regulation is literally: "a rule or directive made and maintained by an authority." I am just scared about who the authority is going to be.

> should be heavily regulated.

By who, exactly? It’s easy to call for regulation when you assume the regulator will conveniently share your worldview. Try the opposite: imagine the person in charge is someone whose opinions make your skin crawl. If you still think regulation beats the status quo, then the call for regulation is warranted, but be ready to face the consequences.

But if picturing that guy running the show feels like a disaster, then let’s be honest: the issue isn’t the absence of regulation, it’s the desire to force the world into your preferred shape. Calling it “regulation” is just a polite veneer over wanting control.

I still believe that even with a reject all button, the regulation is absolute bullshit. The sheer fact that the regulation forbids that setting to be done at the browser level (the law specifically mentions that the consent - or rejection - has to be specific for each website and thus cannot be "once and forget"), is absolute dogshit.

I turn out to be the CTO of a mid-sized SAAS business (100+ FTE over 3 continents), and neither I, any legal advisor, or any other C-level that I have ever met are discussing "what's the fine ? should we just pay instead of complying ?" when dealing with legal stuff. I do not know who brainwashed you into thinking that all businesses are just gonna disregard any regulation given the chance but that has not been my experience at all. I am not saying that none are doing it, but this really isn't the norm. The whole "people = good, business = bad" is just some kind of cozy metal construction that fits in a TikTok video, but that not how the world works.

I m not sure I follow your logic; are you saying that the regulation is not that bad because you are not fined enough if you don't follow it ? Some of us just follow regulations because it's the law - regardless of the fine. I feel like we should be allowed to express our opinion about their merits or shortcomings without considering the penalty aspect which is an entirely separate conversation.

They represent the political opinion of people who are affiliated with them. Once you getting involved in organizations that have a clear and defined political agenda; your whole argument that "nothing you do would ever be politically oriented" and that you are "fully neutral in all situation" becomes incredibly weak.

I am sure some "juge d'instruction" try their very best to be as neutral as possible. Some ostensibly aren't even giving this a flying fuck but both are repeating the same "we are non-political" any time they get the chance. When I hear this, I am unable to know if the person if of the first kind or the second kind. There seems to be 0 investigation internally to weed out the liars which thus casts shadow on the entire profession.

Trust is hard-earned, easily lost, and difficult to reestablish. This scandal touched the very essence of the French judicial system, yet had no major repercussion on the internal organization and processes of those "Juges d'Instruction". It's just business as usual. So until they come up with new systems to ensure better attempt at neutrality and they remove the people that have obviously been plaguing the system for years, it's normal and healthy that any mention of "neutrality" is immediately met with heavy skepticism.

Usually when someone comes with that argument, I ask them to pick any week date in the past year and then I take a random item on my calendar on that day; I give them the time and address of where I need to be as well as the address of my home and I ask them how long it's going to take me and how much it's going to cost. That's usually enough to bring them down a notch from "train work" to "sometimes train work". (But they tend to forget very often, they need to be reminded regularly for some reason). Do you want to play that game with me to get your reality check in order ?

Western Europe in a VERY dense city BTW.

> Magistrats is a broad term that also include Procureurs, Judges but also some Lawyers.

The also is key: "Juge d'instructions" absolutely are "Magistrats" - just like Procureurs, etc are. Some of those "Juge d'instructions" are part of this union who put a target on the back of some politicians. How can they claim with a straight face that they are not biased ?

Either they know it's bullshit and they are simply lying; or they really believe their claims and they are just delusional. I don't know which one I prefer.

Why doesn't the EU do it ?

It's also worth noting that members of the second system had his picture pinned on a wall called "The wall of the assholes"[1] amongst other political and public servant they did not like. They still claim they are totally independent and impartial when judging any of these figures.

[1] (French Wikipedia article about the affair) https://fr.wikipedia.org/wiki/Affaire_du_%C2%AB_Mur_des_cons...

Please ask the EU to lead by example then. The official EU commission website has a cookie banner (https://commission.europa.eu/index_fr)

So either: The EU commission is including trackers on their websites. And they should stop OR they acknowledge that it's almost impossible to build a website without some form of tracking that falls under the law, and they should look into the law itself.

So they have work on their plate.

The author seems very set on following "the proper way of doing things in the linux ecosystem". If I remember correctly, a key principle from Linus himself is: "Talk is cheap. Show me the code". So did the author open any PR to fix any of the issues he surfaced ?

So now you can disregard the notion of "private function" if you pass 100k stars on GitHub ?

Not sure I get your point here