aubanel

Really impressive signal that this 128B model can beat DeepSeek V4 (1.6T) on most coding benchmarks!

Also, I really like Poolside's habit to compare not only to other top models in its weight class (others don't do it, looking at you Mistral), but also to the very top open-weight models, even much bigger ones like the 2.5T Kimi-K3!

Feed algorithms are far from sorting videos only by strict decreasing order of trending factor. One example of that is that feeds are randomized: refresh the page, it changes completely. Another example is how it optimizes for bait content.

Probably a great decision, but why/how can it be decided at a local level by a mayor, instead of a federal level?

This is the absolutely horrific next stage for social media platforms:

- They're already well able to surface the most addictive short video for a specific user out of millions of real videos.

- But these millions of real videos are just darts thrown into the space of "videos that could hook the user", in the end even the best-selected of them is not perfect.

- Now, behold! AI allows to generate the perfect video to surgically hit all the switches in the viewer's brain and turn it into a zombie hooked for days on end.

Let's hope our regulations hit these "social networks" hard enough so that never dare deploy this kind of technology.

Garry Tan's point still stands: he never pretended to be building nice software. But his point was that he can now build AT ALL! Shipping a webpage at all is the firs step ; making it load under 7 Mo is just a refinement, an important one of course (who tf wants bloated webpages) but still only a refinement. Tan is right to be amazed and to be shipping.

There's no question to me, after trying both, that Fable is much better than GLM-5.2 when left alone in front of hard coding tasks Now maybe what plateaus is the human collaboration efficiency, because at some point it will be bottlenecked by the human

Thus companies who still try to have humans perform intertwined work with their AI won't see an improvement, while the ones who fin the right conditions to give their AI more free rein will see it.

Kind of like it's no use having a workhorse pull a combine harvester : at some point, when machines reach sufficient efficiency, you just give wheels to the harvester and let it run.

This is actually a good idea! And still looking forward for aleph alpha to release new models, after the Previa ones a while ago!

Bigger is not better

The article uses the example of GLM being smaller than DeepSeek, yet better on hallucinations as "smaller can be good too"

But the GLM family itself is scaling up fast: GLM-5.x family is 754B, double the previous generation of GLM-4.x

comes within just 4 points of GPT-5.5 and 9 points of Fable 5

9 percentage points IS a big difference

Gotta love how Apple cares (or used to care) about making their products a holistically good user experience

Raw feedback to the team: 1-model looks awesome, 2-The artificially smoothed scrolling on your page feels really bad!

Cheat code to get >5 stars in GTA6 instantly: type "sizes the means of production" in chat

"United Launch Alliance (ULA) is an American launch service provider formed in December 2006 as a joint venture between Lockheed Martin Space and Boeing Defense, Space & Security."

for those who wondered like me!

With what would I be coping? I'd much prefer (probably like most people) if AI were not that powerful. The harsh reality, and the stuff of cope, is that it's (too) powerful.

This is AI-generated, but people upvote it despite its low quality because it generally rubs in the sense of "AI will lose"

This contrast is a bit sad. When Eric Schmidt told students the truth about the importance that AI will take in the future ("It will touch every profession, every lab..."), students booked him But the takes like "AI is not real/powerful, human intelligence is better", which are basically pleasant myopic lies, are cheered. Cope bias is powerful.

And what was your contribution to those achievements to justify this pride?

Of course personal contribution is a factor of pride, and arguably the most justified one.

But it's far from the only one. - fan clubs - a child marvelling on how strong/cool their parents are - US citizens on 4th of July (I'm not American btw)

All of these contributed ~nothing in the phenomenon; their pride comes from the wonders worked by the group they belong to. One does not need to _earn_ pride.

Think it the other way : if you don't think legitimate for the receivers of wonders to feel pride, think of it from the side of the providers of wonders. Parents who toiled for their children, great statespeople who worked hard to improve their country: they intentionally directed their efforts towards someone (descendants, citizens). I think pride is sort of gratitude of receivers for the fruits of a common group's efforts. And it's completely justified IMO to feel un-earned pride.

Looks like the author wants to put on trial all of Railway, Cursor, and even their LLM.

At some point, the responsibility for approving actions made by autoregressive token generations has to belong to the person heading the engineering org... that's you, author.

Does jj work well with parallel agents?

The current problem that I often have is that I want to work on several things in parallel through several agents, always forget to do worktrees, then the different branches of work tend to step on each other Does JJ make it simpler?

A fan of niche medieval history might have threatened the pope with an Outrage of Anagni, much cooler reference than Avignon

That was a cooperation, both sides benefitted. So there's no debt to repay.

I wouldn't be too harsh, scaling x10 YoY is a bit hard on the infra!

This tweet was just reusing an earlier tweet (same image) without attribution : https://x.com/andimarafioti/status/2036107240420032874

Update the link if possible?

I think I have one explanation why for a website, exposing an MCP servers AND having captchas can make sense.

- an agent loading the real page is waste for the server, because the data sent is a few megavytes, and you don't have the usual returns of an user seeing your ads

- BUT API requests (or here, MCP) are much lighter, a few dozen kB, so that makes the ROI positive again

At least that's my view : please tell me, anyone, if that reason doesn't make sense!

Regulating something they visibly had no clue about, just because they had idle time and paper: Is California trying to speedrun the innovation no man's land of EU?

Comments from newly registered accounts on HN are also more likely to mention AI and LLMs

-> to be fair there must also be a bias of young incoming ppl on HN being more prone to be starting their career on the hot new tech

That's a bit harsh! I go to mass every Sunday (in France) and rarely have political stuff. When there, it's most often about abortion or euthanasia (of course in a pro-life (or anti-choice) direction, "you shall not kill")

But dull, empty homilies are (alas) very frequent.

FYI: Chinese models, to be approved by the regulator, have to go through a harness of questions, which of course include this Tiananmen one, and have to answer certain things. I think that on top of that, the live versions have "safeguards" to double check if they comply, thus the freezing.

"well preserved tools" said the ad -> I bought some, surprisingly expensive for a hammer -> it's a mishap and inform piece of wood -> straight to dump

Yet the carbone intensity of energy production in Germany is among the worst in Europe.

And France (nuclear powered, no particular huge investment in a green transition) beats them easily in both price and carbon.

https://app.electricitymaps.com/map/live/fifteen_minutes

The success of the Energiewende ? Just take a look at average electricity price and carbon footprint in Germany Vs France.

https://app.electricitymaps.com/map/live/fifteen_minutes