Pinned Tweet
Rejoice 16 GB M6 Mac mini owners 🍎
Ternary Bonsai 2 27B now runs on the Neural Engine:
- 52 tok/s on code (GPU: 36)
- 300 tok/s prefill (GPU: 185)
- 12 W (GPU: 30 W)
- Works with Grok Build
Reproduced on a second M6 mini.
github.com/cowboycoderhq/Ter…
Cowboy Coder retweeted
To my American friends, part two. Last time I talked about France. This time I want to talk about something bigger than France, bigger than any midterm you've ever voted in: who will control artificial intelligence.
You carry a responsibility no other people on earth carries right now. The most powerful technology in human history is being built in your country, under your laws. Whatever rules you choose will become the rules for the world. If Democrats retake Congress, here is the playbook I see coming, because in Europe we've already watched the first chapters.
It starts with a beautiful promise. Superintelligence will bring abundance. Cures for every disease, free education, the end of scarcity. And it might. But then comes the sentence that always follows: "This is too important to be left to the market."
Then comes regulation. It won't be one law. It will be a thicket of them: licenses to train a model, mandatory audits, compliance offices, approved datasets, safety boards staffed by people who have never shipped anything. Each rule sounds reasonable on its own. Together they build a wall.
Who can climb that wall? Not the kid in a garage in Ohio. Not the two-person startup. Only the giants with armies of lawyers and lobbyists. Some of the biggest AI companies have already gone to Congress asking for a licensing regime. When incumbents beg to be regulated, it's not because they love rules. It's because rules are a moat.
So you end up with two or three approved AI companies sitting next to the government, dressed up as competition. They would share the same values, the same filters, and the same red lines, all decided in the same rooms. Europe just did this with the AI Act. The result is that we barely build anything. We regulate what Americans and Chinese create.
Now think about what that means once AI is everywhere. It will decide what information you see, what loans you get, which jobs you're screened for, and what speech gets flagged. A small class of officials and their partner corporations would hold the master switch on human knowledge. That is a new nomenklatura. The Soviet apparatchiks needed millions of informants to control what people thought. This one would need a few data centers and a compliance policy. Orwell imagined a telescreen in every room. We're building something far more intimate.
This is exactly what your Founding Fathers feared. They had just escaped a distant power that decided everything for them, so they designed a country where power could never concentrate in one place. They split it, checked it, and handed the most important guarantees directly to the people: the right to speak and the right to bear arms. Those rights were never gifts from the state. They were limits on the state. Free speech came back to life the day X stopped being a censorship machine. But a Constitution is only as strong as the citizens who defend it at the ballot box. If AI gets centralized in a handful of approved hands, those rights become words on old paper.
Have you watched Altered Carbon? In that world, the ultra-rich live for centuries by buying new bodies. Everyone else gets one life, rents their existence, and dies. Society hasn't collapsed. It works perfectly for the people at the top. That's the real horror: it's stable. No amount of work, talent or merit can lift someone from the bottom, because the ladder itself is owned. That's what a captured AI economy looks like. It won't be chaos. It will be a permanent ceiling.
America has always been the opposite: the one country on earth where a nobody with a laptop can become somebody. I build my company between Paris and San Francisco. I know which side of the Atlantic still lets you try.
So before you vote, ask every candidate one simple question: will your AI policy let a 19-year-old in a dorm room compete with the giants, or will it protect the giants from that 19-year-old?
Keep AI open. Keep power distributed. Keep the ladder.
Stay American.
I’m French. I want to talk to Americans who are about to vote Democrat. Not to insult you. To warn you.
I come from your future.
In France we did everything your progressive politicians promise. Free healthcare, free college, generous welfare, strong unions, heavy regulation, high taxes on the rich, a state that takes care of everything. Public spending close to 57% of GDP, one of the highest on the planet. On paper, it’s the dream.
Here’s what it looks like from the inside. France hasn’t passed a balanced budget since 1974. Our debt is above 3,300 billion euros. Interest on that debt is becoming one of the biggest items in the national budget. Our hospitals are collapsing despite record funding. Our schools rank lower every year. Young people with degrees can’t afford an apartment in the city they were born in. Our best engineers leave for London, Zurich and San Francisco. And the rich you wanted to tax? They left too, and took the jobs with them.
The promise was protection. The result is stagnation.
The worst part is that nobody in France voted for decline. Every single step sounded generous, compassionate, reasonable. One more program. One more regulation. One more tax to fund it. Each step alone makes sense. Fifty years of them, and you wake up in a country that can no longer build a nuclear plant on time or keep its streets safe.
I love America because it’s the one place on earth where people still believe they can build something from nothing. I came to Silicon Valley as a self-taught kid and it changed my life. That spirit is the most precious thing your country has.
Before you vote, ask yourself one question: is the program you’re voting for making America more like America, or more like France?
Trust me. You don’t want to be France.
Cowboy Coder retweeted
Elon Musk often says: "Patents are for the weak."
Why does he say that? In 2014, Tesla opened up all its patents. Anyone could use them. Everyone thought it was a PR stunt. It was actually a declaration of war. Elon's point was that patents mostly serve to slow down progress, protect the positions of giant corporations, and make lawyers rich. Not to innovate.
His logic is simple: if your competitive advantage lives inside a legal document, you're already dead. The only real moat is your speed of innovation. If you move faster than everyone else, you don't need patents. If you need patents to survive, it means you've stopped moving fast enough.
The strong build. The weak file.
And AI is about to make that sentence true for the entire economy.
Take biotech. For 50 years, starting a biotech company meant a PhD, ten years in a lab, access to closely guarded knowledge, millions before your first proof of concept, and a network in the right academic and regulatory circles. Knowledge was scarce, access was scarce, and that scarcity was worth billions.
That scarcity is dying.
I genuinely believe we're entering a world where anyone will be able to start a biotech company. A 19-year-old with AI agents will be able to explore molecular spaces that entire labs used to spend years mapping. Trade secrets, credential rents, access rents, positional rents: all of it is about to blow up.
And look at what's already happening. Some of the first calls to "urgently regulate AI agents" are coming from established biotech players. Coincidence? When your business model depends on scarce knowledge and knowledge becomes free, you're left with only two weapons: the patent and the regulator.
The two weapons of the weak.
Here's what's coming. A wave of panic. Dozens of labs and incumbents retreating behind their patents, filing lawsuits, and above all pouring money into regulatory capture. They won't try to build better. They'll try to make it illegal for anyone else to build.
Because the truth is, a lot of people in business were never strong. They were protected. They never won on product, on innovation, on speed of execution. They won because the market was locked. Because attacking their industry required hundreds of millions of dollars, years of approvals, and armies of lawyers. The wall was never their talent. The wall was regulation.
And when the wall comes down, we find out who actually knew how to build.
This isn't just biotech. Banking. Insurance. Law. Healthcare. Energy. Hardware. Education. Defense. Almost every industry where incumbents were sleeping behind barriers to entry is about to get hit by thousands of founders able to attack their business from every angle at once. With teams of 5 people and an army of agents.
Everyone gets to play now.
My conviction, and I'll say it plainly: a huge chunk of the S&P 500 could be completely flushed in the next 5 years. Not because those companies will make mistakes. Because their competitive advantage was never real. It was built on how hard it was to get in, and that difficulty is disappearing.
The weak will cry danger. They'll talk about safety, risk, disaster. Sometimes they'll be right on a specific point. But most of the time, what they're protecting isn't the public. It's their rent.
What's coming isn't a collapse. It's a renaissance. The biggest explosion of value creation in history, driven by people who were always told they weren't allowed to play.
Patents are for the weak.
And the weak are about to have a very bad decade.
They are afraid. AI will disrupt and "open source" everything, not just games and software. It will also go after almost all hardware and drugs.
Basically everything.
Some things sooner, some later.
Everything will change. Current ways of doing things must/will change.
Economy will look very different by 2030s.
Cowboy Coder retweeted
Je veux présenter mes excuses, au nom des Français, pour avoir enfanté la French Theory (qui a enfanté la pire des merdes idéologiques : le wokisme).
Nous avons donné au monde Descartes, Pascal, Tocqueville. Et puis, dans les ruines intellectuelles de l'après-68, nous avons donné Foucault, Derrida, Deleuze. Trois hommes brillants qui ont fabriqué, dans l'élégance de notre langue, l'arme idéologique qui paralyse aujourd'hui l'Occident.
Il faut comprendre ce qu'ils ont fait. Foucault a enseigné que la vérité n'existe pas, qu'il n'y a que des rapports de pouvoir déguisés en savoir. Que la science, la raison, la justice, l'institution médicale, l'école, la prison, la sexualité, tout n'est qu'une mise en scène de la domination. Derrida a enseigné que les textes n'ont pas de sens stable, que tout signifiant glisse, que toute lecture est une trahison, que l'auteur est mort et que le lecteur règne. Deleuze a enseigné qu'il fallait préférer le rhizome à l'arbre, le nomade au sédentaire, le désir à la loi, le devenir à l'être, la différence à l'identité.
Pris isolément, ce sont des thèses discutables. Combinées, exportées, vulgarisées, elles forment un système. Et ce système est un poison.
Car voici ce qui s'est passé. Ces textes, illisibles en France, ont traversé l'Atlantique. Les départements de Yale, de Berkeley, de Columbia les ont absorbés dans les années 80. Ils y ont trouvé un terreau qui n'existait pas chez nous : le puritanisme américain, sa culpabilité raciale, son obsession identitaire. La French Theory s'est mariée à ce substrat, et l'enfant de ce mariage s'appelle le wokisme.
Judith Butler lit Foucault et invente le genre performatif. Edward Said lit Foucault et invente le post-colonialisme académique. Kimberlé Crenshaw hérite du cadre et invente l'intersectionnalité. À chaque étape, la matrice est française : il n'y a pas de vérité, il n'y a que du pouvoir, donc toute hiérarchie est suspecte, toute institution est oppressive, toute norme est violence, toute identité est construite donc négociable, toute majorité est coupable.
Voilà comment trois philosophes parisiens, qui n'ont probablement jamais imaginé leurs conséquences pratiques, ont fourni le logiciel d'exploitation à une génération entière d'activistes, de bureaucrates universitaires, de DRH, de journalistes, de législateurs. Voilà comment on a obtenu une civilisation qui ne sait plus dire si une femme est une femme, si sa propre histoire mérite d'être défendue, si le mérite existe, si la vérité se distingue de l'opinion.
C'est de la merde pour une raison simple, et il faut la dire calmement. Une civilisation se tient debout sur trois piliers : la croyance qu'il existe une vérité accessible à la raison, la croyance qu'il existe un bien distinct du mal, la croyance qu'il existe un héritage à transmettre. La French Theory a entrepris de dynamiter les trois. Pas par méchanceté. Par jeu intellectuel, par fascination du soupçon, par haine de la bourgeoisie qui les avait nourris. Mais le résultat est là . Une génération entière a appris à déconstruire et n'a jamais appris à construire. Une génération entière sait soupçonner et ne sait plus admirer. Une génération entière voit le pouvoir partout et la beauté nulle part.
Je m'excuse parce que nous, Français, avons une responsabilité particulière. C'est notre langue, nos universités, nos éditeurs, notre prestige qui ont donné à ce nihilisme son emballage chic. Sans la légitimité de la Sorbonne et de Vincennes, ces idées n'auraient jamais traversé l'océan. Nous avons exporté le doute comme d'autres exportent des armes.
Ce qui se construit maintenant, en silicon valley, dans les labos d'IA, dans les startups, dans les ateliers, dans tous les lieux où des gens fabriquent encore des choses au lieu de les déconstruire, c'est la réponse. Une civilisation se reconstruit par les bâtisseurs, pas par les commentateurs. Par ceux qui croient que la vérité existe et qu'elle vaut qu'on s'y consacre. Par ceux qui assument une hiérarchie du beau, du vrai, du bon, et qui n'ont pas honte de la transmettre.
Alors pardon. Et au travail.
Cowboy Coder retweeted
Encore une fois, Elon Musk avait raison. Et les scientifiques vont se prendre une vraie blessure narcissique.
Il y a deux mois, j'ai dîné avec un professeur de physique quantique à Normale Sup, l'une des meilleures écoles de France.
Il se présente avec cette phrase :
"Je suis le seul prof pas gauchiste dans un marécage de profs gauchistes décroissantistes."
Ses collègues sont tous anti-IA. La raison ? Ça consomme de l'eau.
Le mec est passionnant. On part très loin : l'espace, le temps, la conscience. Des discussions complètement perchées.
Et à un moment, il me lâche :
"Tu sais, depuis les années 70, aucun sujet vraiment sérieux en physique n'est financé."
Je lui demande ce qui est financé, alors.
"Des cathédrales théoriques. Magnifiques. Qui ne servent quasiment à rien."
Je lui demande pourquoi.
"Après tout ce qui s'est passé avec le nucléaire, le monde académique s'est imposé un principe de précaution. Il a peur de trouver des choses réellement profondes."
Relisez cette phrase.
La science a peur de trouver.
C'est exactement ce que décrit François Chollet aujourd'hui : la théorie des cordes, c'est ce qui arrive quand une intelligence est coupée du réel. Elle hallucine. Elle construit des cathédrales. Elon a répondu avec un seul emoji : 🎯
Ce n'est pas un accident. C'est une capture.
Je parle souvent de Gramsci. Son idée centrale : on ne prend pas le pouvoir par la force, on prend les institutions. Universités, médias, culture. Et une fois qu'on les tient, on décide de ce qui est pensable.
La recherche n'y a pas échappé. On finance ce qui est publiable, consensuel, sans risque. On ne finance plus ce qui pourrait tout changer.
C'est pour ça que beaucoup de gens en France pleurent en ce moment. Ils sentent que leur monde factice est en train de mourir. Ce monde où un titre suffisait. Où un statut remplaçait un résultat. Où l'on pouvait passer trente ans à écrire des papiers lus par douze personnes sans jamais se confronter au réel.
Elon l'a dit sans filtre : une grande partie des PhD ne servent à rien. Et chez xAI, il a supprimé le titre de "chercheur". Il n'y a que des ingénieurs.
Le message est limpide. On ne te paie pas pour chercher. On te paie pour trouver.
La science va devoir retrouver son vrai but. Pas produire des papiers. Pas décrocher des grants. Pas briller en colloque. Comprendre le réel, et le plier.
Et l'IA va accélérer tout ça brutalement. Quand l'intelligence devient abondante, elle n'est plus le goulot d'étranglement. Ce qui compte, c'est le courage de la confronter au monde physique, encore et encore.
Alors physiciens, mathématiciens : il est temps de vous sortir les doigts du cul et de créer de la vraie valeur pour le monde.
On veut l'antimatière.
On veut l'antigravité.
On veut le warp drive.
On veut comprendre la conscience au niveau de l'infiniment petit.
Les vrais sujets. Ceux qu'on a eu peur de toucher pendant cinquante ans.
La science a été capturée. Elle va être libérée.
Et ça va faire très mal à certains egos.
Same Mac, same prompts, each engine alone, 3 runs:
llama.cpp: 71 prefill / 18 decode / 35 W
mlx-serve: 185 / 17 / 20 W
mlx-serve + DFlash2: 183 / 36 / 30 W
ANE (this): 301 / 52 / 12 W
Decode on code.
Asteroids benchmark demo:
cowboycoderhq.github.io/Tern…
Cowboy Coder retweeted
Replying to @TheDavidTai
@TheDavidTai did amazing work here. This is worth reading.
Coming very soon to MLX-Serve… next release will be so big. Definitely doing a “preview” release first, so please look out for that on the GH release page.
DS 4.1 for 128 machines that is useable.
Race will be split until someone figures out how to utilize bandwidth for prefill. I feel like the excess bandwidth could be utilized to simultaneously employ the same, but so far haven’t built anything concrete. Anyone got a solution?
Mark my words, the mlx community is awakening. What holds back local Super Intelligence on Apple silicon isn’t the hardware, it’s the software.
MLX-Serve 26.10.1 is out: speed across the board.
Exactly faster. Qwen3.8 27B with its drafter
+66% on an M5 Ultra,
+28% on an M4 Max,
+37% on an M1 Pro.
Qwen Flash Next:
+11% decode on M4 Max,
+51% prefill on M5 Ultra. (up to ~5400 PP!!!!)
Gemma 4, Qwen, LFM2.5, Spark, Bonsai 2 all faster, output byte-identical to 26.9.6 (across 18 models).
New:
* GGUF models run on our own MLX engine, written in Zig and Metal, instead of the llama.cpp fallback. (experimental)
* 🍣 Sushi-format Flash Next packs run directly.
* Qwen-Image 2.1 edits pictures from instructions.
* Drafting is on by default for every model that supports it.
Huge thanks to everyone who filed, tested and fixed this one: @STRML_, @sbusso, @wutang_superfan , lborloz, LXD-8 @alinselea zeeshanhaque21 @Lojza3D, h9q2cyxvgm-ui @kennethrdegraff, @CowboyCoderHQ , codysk, @Beamsters1, @CerebralCoding_, @AjAbsaki , and more.. THANKS ! Please comment if I did not include your name, for visibility, I know people mostly by their GH handle.
Special thanks to @ashxhart and @Spangler3000 for pushing !
Benchmarked and tested across 5 machines: htmlhost.jax.workers.dev/ren…
GH Release Download + Changelog:
github.com/ddalcu/mlx-serve/…
Website: mlxserve.com
Cowboy Coder retweeted
MLX-Serve 26.10.1 is out: speed across the board.
Exactly faster. Qwen3.8 27B with its drafter
+66% on an M5 Ultra,
+28% on an M4 Max,
+37% on an M1 Pro.
Qwen Flash Next:
+11% decode on M4 Max,
+51% prefill on M5 Ultra. (up to ~5400 PP!!!!)
Gemma 4, Qwen, LFM2.5, Spark, Bonsai 2 all faster, output byte-identical to 26.9.6 (across 18 models).
New:
* GGUF models run on our own MLX engine, written in Zig and Metal, instead of the llama.cpp fallback. (experimental)
* 🍣 Sushi-format Flash Next packs run directly.
* Qwen-Image 2.1 edits pictures from instructions.
* Drafting is on by default for every model that supports it.
Huge thanks to everyone who filed, tested and fixed this one: @STRML_, @sbusso, @wutang_superfan , lborloz, LXD-8 @alinselea zeeshanhaque21 @Lojza3D, h9q2cyxvgm-ui @kennethrdegraff, @CowboyCoderHQ , codysk, @Beamsters1, @CerebralCoding_, @AjAbsaki , and more.. THANKS ! Please comment if I did not include your name, for visibility, I know people mostly by their GH handle.
Special thanks to @ashxhart and @Spangler3000 for pushing !
Benchmarked and tested across 5 machines: htmlhost.jax.workers.dev/ren…
GH Release Download + Changelog:
github.com/ddalcu/mlx-serve/…
Website: mlxserve.com















