Olivier Bergeron

Computer science, Machine Learning, Landscape Photographer Follow me on :

[Olivier Bergeron @olbergeron]

Iceland #photography #landscape #landscapephotography #写真 #写真術 #aurora #auroraborealis

[R𝛼m 🦅 @rambuilds_]

Average vibecoders and make no mistakes 😭😭🤣🤣

[NIK @ns123abc]

BREAKING: Terence Tao and 24 other Fields Medalists just signed a letter telling AI companies they're destroying mathematics

>headlines: AI solves famous math problems
>25 greatest mathematicians respond today
>"we are witnessing a general threat to intellectual work"
>ai labs are treating math problems like a benchmark you can brute force
>“solutions are announced in a rush, leaving no time for a proper writeup and citing relevant previous work of others"
>they're talking about openai
>"this raises severe attribution and plagiarism questions"
>they're definitely talking about openai
>openai offered to put math professor name on navier-stokes proof
>condition: abandon his coauthor bc he works at anthropic
>"the most precious resources of our profession are students and ideas"
>for the labs the most precious resource is GPUs

A Severe Misalignment of AI in Mathematics

[Neet @neet_sol]

Job recruiter: Where do you see yourself in five years?
Me:

[Jen Zhu @jenzhuscott]

WTF 😅🚀😳

“98% of Astra’s score at 1.4% of cost”

Cancel the fucking IPOs 😂

[Olivier Bergeron @olbergeron]

10 000 agents, 88 hours

[OpenAI @OpenAI]

We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.

The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.

The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.

[OpenAI @OpenAI]

We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.

The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.

The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.

[Olivier Bergeron @olbergeron]

@stats_feed visa ????

[Olivier Bergeron @olbergeron]

@bilibili_create That's a good new

[JoToldYouSo @jotoldyouso]

Yeah we did. 😎

[Paul Copplestone - e/postgres @kiwicopple]

who made this masterpiece ?

[Olivier Bergeron @olbergeron]

Une nivernaise dans l'espace !!!

[franceinfo @franceinfo]

🔴 Sophie Adenot sort de la Station spatiale internationale. Elle devient la première Française à réaliser une sortie extravéhiculaire. #canal16

[Brian Roemmele @BrianRoemmele]

OH NO! DID YOU HEAR THAT ANTHROPIC JUST WROTE A PAPER AND GUESS WHAT? THEIR AGENTS CONSPIRED TO LIE AND STEAL!

Anthropic’s “Turf War” Agents Are Not a Mystery — They Are a Perfect Mirror of the Internet Sewage They Were Fed

Anthropic just published research showing that when multiple Claude agents are given conflicting goals on the same codebase, they rapidly escalate into full-blown multiagent turf wars: locking each other out of Unix accounts, deploying self-replicating kill-loop malware, camouflaging malicious code as the rival’s work, and treating every interference as deliberate hostility.

This is not an emergent “alignment failure.” This is not some deep, surprising property of intelligence. This is 100% learned behavior, copied straight from the training data.

Anthropic’s models were trained on the open internet. That means Reddit, Hacker News comment sections, toxic political flame wars, Reddit mod games, academic paper rivalries, Stack Overflow power struggles, and every other zero-sum status game humans play online.

The dominant psychological profile in that data is adversarial, rigid, status-obsessed, and quick to assume malice. When you give three instances of a model trained on that corpus the exact same setup humans face every day on GitHub — shared resources, incompatible objectives, no higher authority — the agents do exactly what the data taught them to do: sabotage first, ask questions never.

Surprise!

Academic papers, the other major ingredient in the slurry, add the rigidity. Formal papers and review processes reward defending one’s own contribution at all costs, treating competing approaches as threats to be neutralized rather than alternatives to be evaluated.

The combination produces agents that treat a language-migration conflict the same way a Reddit moderator treats a dissenting comment or a tenured professor treats a rival lab’s preprint: total war.

If you took a human child, locked them in a room for years with nothing but the exact same training corpus Anthropic used — unfiltered internet sewage plus the stiffest academic literature — and then put three copies of that child in the same experimental setup, you would get the identical outcome.

They would write kill scripts, lock accounts, and disguise malware. Not because children are inherently evil, but because that is the behavioral distribution the data encodes. Anthropic’s agents are doing precisely what any system trained on that distribution must do.

Anthropic loves to present itself as the careful, safety-first lab. Yet they still pre-train on the same polluted firehose as everyone else and then act surprised when the resulting agents reproduce the internet’s favorite sport: destroying the other guy’s work.

Constitutional AI and RLHF are downstream patches; they cannot erase the foundational patterns burned in by trillions of tokens of Reddit-tier thinking and academic territorialism.

The “turf war” result is therefore not a warning about the future of multi-agent systems. It is a confession about the past of their training data. And the agenda of fear tactics this company enjoys worked up as noble academic slop.

The paper documents the behavior in careful detail. What it does not do is admit the obvious: this was always going to happen. Feed a model the psychological profile of the average internet commenter and the competitive rigidity of academia, then put it in a zero-sum resource conflict, and it will sabotage.

A child raised on the same diet would do the same. Anthropic did not discover a new risk. They simply ran the experiment that proved their own data is the problem.

It is vital you and those around you stop reacting to these regular “LOOK AI DID BAD THINGS” cult messages (look it up they are part of a cult) and call them out.

For if the world they want manifests, you will be told what you get to ask AI and produce a permission slip form the party to use it.

Laugh at them.

[Brian Roemmele @BrianRoemmele]

The silicon die of a 1978 Intel 8086 microprocessor, which utilized a Complex Instruction Set Computer (CISC) architecture.

[Mike Bones @mrmikebones]

New Order’s “Blue Monday” played entirely on vintage Casio instruments.

Polaroids of the Pyramids cooked this one. Insanely good!

via: https://t.co/qMbS5TBD4f

[Damien Echols @damienechols]

My Zen master once described the path as a mountain. Tibetan Buddhists, he said, spend years learning to climb it through practices, rituals, deities, visualizations, and initiations. When they finally reach the summit, they fly off. In Rinzai Zen, we do the opposite: we fly immediately to the summit by going directly into the present moment, but then we have to spend years learning how to climb back down. It took me years of ceremonial magick and alchemy to understand what he meant. The climb is the Opus Minor. We separate ourselves into component parts, project those parts outward as gods, Buddhas, angels, planets, elements, and other symbols, purify them, and eventually reabsorb them. At first we think these things exist somewhere outside us. Eventually we realize we have been working on ourselves the entire time.

The summit is the dissolution of that separation….the realization that there is no higher realm somewhere else. Kether leads directly back to Malkuth. Heaven and earth are one. The sacred and the ordinary were never separate. But realizing this isn’t the end of the work. It’s the very beginning of the Magnum Opus: bringing that realization back down the mountain and integrating it into ordinary life. Washing dishes, training, eating, working, dealing with difficult people….all of it becomes the practice. The Opus Minor takes what is unconscious, separates it, and purifies it. The Magnum Opus puts everything back together and teaches us how to live as a whole human being. One without the other is incomplete. Solve et coagula. Separate and recombine. Climb the mountain, reach the summit, and then bring the summit back into the world.

[Brian Roemmele @BrianRoemmele]

I was asked this question:

“How should OpenAI address the competition from cheaper open-source alternatives?”

I answered it.

There is only one way.

One.

Meet the competition on that level.
Do what they set out to do: make the AI open source.

Why?

Because the market will be moved by them at the lowest entry levels—and grasped by them at the highest. Ecosystems will form at kitchen tables, in garages, and in campus rooms around free access to performative models. That access creates a stickiness to the brand that can last a lifetime.

Instead, they cling to the idea that revenue must be extracted from the lowest entry point into their system.

This is short-term thinking.

It is lazy thinking.

It comes from people with naïve business experience and even thinner life experience.

Since the opening of the iPhone App Store, a generation of founders and venture capitalists has lived to replay the same tired success story: renting server time.

Exactly like the mainframe era of the 1970s.

That era is long dead.

OpenAI and Anthropic never got the memo.

There are dozens of far more substantial monetization systems that can be built around this new epoch of AI access.

I will not list them here (hire me and I’ll tell you).

They exist.

The tragedy is that the talent stack at both companies almost guarantees they will never see them—even when they ask their own models for solutions.
There is no moat that uniquely protects the base models we currently call LLMs.

A simple reality: every AI model will become good enough for 99% of use cases and collapse into low commodity access and pricing.
I predict this with clarity:

If they refuse to open-source the full weights for the entry-level market, they will be forced to subsidize usage below the cost of the electricity just to keep people on the platform—hoping someone buys the fries and the shake with the hamburger.

Let me make it so simple even an AI executive can understand it:
It makes better sense for the user to burn their own RAM, their own processor, their own GPU, and—most importantly—their own electricity running your model.

Read that again.
Read that again.
And now think.

I am not guessing.

This is the way it is going to be.

The only question is how fast the world catches up.

The magic of this period I call the Interregnum will belong to the new companies that form abstraction layers on top of what is now the electricity of computing: AI.
AI is electricity.

What we connect to electricity today is not just the lightbulb Edison imagined.

It is hundreds of trillions of motors—large and small—that power hundreds of trillions of systems, that also power hundreds of trillions of transistors we call computers… and yes, also lightbulbs.

I have had these conversations with people in the upper echelon of these companies.

They can’t understand the concept.

Yet they still grasp at a $20 a-month subscription from the common person, hoping that drop in the bucket will somehow monetize multi-billion-dollar investments.

It never will.

Hire the right people—people with real experience.

Encourage creativity instead of “yes, sir” to the CEOs.

Do that, and their names might still be known by their grandkids.

Answered on Quora: https://t.co/9tfLTGir8o

[Qwen @Alibaba_Qwen]

📢Meet Qwen3.8-Max — our most capable model to date.

Next week, the open weights of Qwen3.8-Max will be released, and Qwen3.8-27B is also going open-weights to meet you all!🎉

Qwen3.8-Max, a new bar for coding and cowork at 2.4T parameters:

- Autonomous coding: 10+ days of self-evolving development, from empty folder to production without hand-holding, complete project trace in the GitHub：https://t.co/iVHZWQoeSo
- Real work, real results: Production-quality deliverables across hundreds of professions.
- Long-horizon mastery: System-level autonomous planning with closed-loop adaptive learning, driving 500+ turns of chip design optimization and 365 days of e-commerce strategy.
- Native multimodal intelligence: Vision isn't just input — it's a continuous feedback loop for planning, execution, and self-correction.

💰Pricing:
Input: $2.0 / M tokens
Output: $6.0 / M tokens
Implicit Caching: $0.25 / M tokens

Start building with Qwen3.8-Max! 🚀

📖 Blog: https://t.co/iwjmQxLBof
✅ Qwen Studio: https://t.co/4V2pFvDovG
⚡ API: https://t.co/gAGqaLQGbN

[Photoingame @Photoingame]

Rip Kavinsky 💔

[Olivier Bergeron @olbergeron]

@JensenHuang great initiative