Manu #fella

@sflorimm no Trump and no Xi. + same access to Claude, Google and chinese oss models

USA has ChatGPT USA has Grok USA has Claude USA has Gemini USA has Llama USA has Copilot China has DeepSeek China has Qwen China has Ernie China has GLM China has Kimi China has MiniMax Europe has?

@boss_nathan @mark_slapinski your comment is stupid.

Strong feelings don't equal strong arguments. "Worst president in history" is a claim that requires comparison to leaders who presided over actual famines, civil wars, or institutional collapse. That's a debatable case, not a settled one. The specific labels pedophile, war criminal are serious legal designations that require evidence and conviction, not political conviction. Throwing them as insults cheapens the weight they carry when applied accurately to actual offenders. You can have a legitimate, well-reasoned critique of Trump's presidency — policies, character, rhetoric, institutional damage. There's real material there. But this isn't critique. This is a rage post dressed up as courage. Losing followers for truth is noble. Losing them for unsubstantiated name-calling is just noise.

I might lose followers for this, but it needs to be said: Trump is BY FAR the worst president in American history. He's a pedophile, a con artist, and a war criminal. May he burn in hell.

@PeterDiamandis supervised* with your active supervision with critical disengagement every 1000 miles. still very very far from being safer than human. you are a liar

Tesla's FSD: 5.3 million miles between accidents. US driving average: 660,000.  That's 9x safer. And it's only getting better.

@MrMikeInvesting this you? x.com/i/status/20145…

Hear me out… These 5 names will create generational wealth for many over these next 9 months. $ASTS is a $250 stock trading at $90. $RKLB is a $160 stock trading at $70. $NBIS is a $220 stock trading at $120. $AMD is a $350 stock trading at $195. $ONDS is a $32 stock trading at $10. The opportunity behind these names is so massive that they’ll be talked about for generations to come. Save this for later…

@cryptopunk7213 cursor ppl didn't write opus

im genuinely pumped for the XAI comeback. they’re throwing 500,000-1M blackwells at these models and hiring aggressively, what do you think happens? its clear coding is the frontier. if your model can't code forget it. grok build uses 8 agents in parallel to write your codebase. they just hired cursors top people to design this. i know elon lost round 1 but im still betting they leapfrog competitors by end of 2026

@Ahead_of_Curve and yet, it's still supervised.

Ever since FSD (Supervised) became available in Australia in August last year, I’ve supervised my car driving me for approximately 15,000 km. In that time, the car would have driven me past thousands of other drivers and passengers on the road, yet I doubt more than a handful—if any—realised it was the Tesla driving itself rather than me. If self-driving software is based on computer code the cars will stick out in traffic as they will feel and behave like a rigid robot but importantly there is no way the code will be sophisticated enough to handle rare edge-case scenarios the real world throws up. By contrast, when the system is trained on billions of miles of real human driving data, it learns to behave and drive like a human. It blends so seamlessly into the flow of traffic that other road users almost never realise the person behind the wheel, while, they may have proper control, is merely supervising. It’s the car that’s actually in control of the full suite of driving inputs including steering, both foot pedals, indicators, high beams when needed, and even operating the windscreen wipers as conditions change. The driver sits ready to intervene, but in everyday driving, the Tesla’s actions feel so natural and human-like that no one gives it a second glance. To achieve safe, seamless and human-like driving the software requires an immense amount of real-world human driven data to learn how to deal with rare situations and edge case scenarios. As at the beginning of the month, Tesla’s Full Self-Driving (Supervised) system had accumulated over 8.4 billion miles (roughly 13.5 billion km), with billions more being added each year. (For perspective that’s about 42 times more than Waymo’s 200 million miles at the same time.) As a result, according to Tesla’s reported US data, the software is already showing significantly fewer major collisions than the average human driver—in the range of being about 7-8 times safer. That enormous training volume—equivalent to circling the Earth’s equator hundreds of thousands of times, or making around 30 round trips to Mars—allows the neural network to absorb nuanced, human-like driving patterns. It masters reactions to rare events and picks up on subtle social cues on the road, so the car’s behaviour feels human-like, natural and predictable rather than robotic or overly cautious. As the data keeps growing (heading towards the 10 billion mile mark that Elon Musk has highlighted as critical for tackling the long tail of edge cases), that human-like seamlessness will only improve. Then, once FSD Unsupervised arrives and people start noticing empty driver’s seats, they’ll suddenly realise just how many self-driving cars are already out there—and why accident rates and road fatalities are beginning to drop.

@elonmusk DOGE was supposed to reduce gov debt. It only caused more debt. Such a failure. your only wins are transphobic That sums it all.

Legendarily based 🦾

@elonmusk @peterwildeford yes sure. and humans on March in 2025.

@peterwildeford xAI will catch up this year and then exceed them all by such a long distance in 3 years that you will need the James Webb telescope to see who is in second place

Based on the data I see, I think: - Anthropic🇺🇸/Google🇺🇸/OpenAI🇺🇸 all ~tied - Meta🇺🇸 / xAI🇺🇸 each ~7mo behind - Moonshot🇨🇳/- Deepseek🇨🇳 / zAI 🇨🇳 / Alibaba🇨🇳each ~9mo behind - Mistral🇫🇷 ~1.5 years behind - No other companies competitive

@haider1 yet nobody uses it for real coding work

xAI is really strong on price to performance grok 4.20 (beta) is opus 4.5 level with significantly faster inference, significantly lower cost, and a lower hallucination rate opus 4.5 costs $5 input and $25 output per million tokens, while grok 4.20 costs $2 and $6

@MarioNawfal @elonmusk looool

Elon: Optimus 3 is almost done. Entering the final stage. Most advanced robot on Earth. Nothing else compares. Production starts this summer. Classic S-curve ramp. High volume next year. Optimus 4 next. New robot. Every year. That’s the goal. @elonmusk

@davidpattersonx Tesla cars will replace the entire car industry. Grok will replace all LLMs Hyperloop will replace... always will, will, will.... but never has, has, has.

Macrohard will replace entire companies doing information work. Optimus humanoid robots will replace all physical work. Both will be guided by Grok AGI.

@brivael si le prix du marché est €50, tu vends pas à €2 mais à €48.

"Non mais moi l'IA ça me concerne pas, ça fait 20 ans que je fais mon métier" Ok. Laisse-moi t'expliquer ce qui arrive. En ce moment, des gens montent ce qu'on appelle des "AI Agencies". Le concept est simple : Tu prends un expert du domaine (un avocat avec 20 ans de barreau, un comptable senior, un consultant stratégie) + une personne qui maîtrise les systèmes d'IA et sait construire des workflows agentique. À deux, ils construisent une machine qui délivre la même prestation qu'un cabinet de 15 personnes. Même qualité. Même rigueur. Le client ne voit aucune différence dans le livrable. La différence ? Le prix. 10% du tarif actuel. Et souvent, plus rapide. Imagine que la baguette coûte 50€ partout depuis toujours. Tout le monde trouve ça normal, c'est le prix du marché. Et là, quelqu'un débarque avec une baguette à 2€ et elle est aussi bonne, voire meilleure. Tu crois que les gens vont continuer à payer 50€ par fidélité ? C'est exactement ce qui va se passer dans le conseil, le juridique, la compta, l'audit, le design, le développement. Partout où il y a de la prestation intellectuelle facturée à l'heure. Ces agences n'ont pas besoin de 50 salariés. Pas de locaux immenses. Pas de juniors à former pendant 3 ans. Elles ont un expert, des outils, et des systèmes. Et le pire : c'est pas dans 15 ans. Les outils existent aujourd'hui. Les premières agences se montent en ce moment. Le cabinet qui facture 500€ de l'heure pour du travail que deux personnes avec les bons systèmes font en 10 fois moins de temps ne pourra pas tenir. Pas par manque de compétence par impossibilité structurelle de rivaliser sur les coûts. La question n'est plus "est-ce que l'IA va toucher mon métier". C'est : est-ce que tu seras celui qui construit la baguette à 2€, ou celui qui essaie encore de vendre la sienne à 50 ?

@DillonLoomis Tesla needs a safety driver. Waymo doesn't.

We now know: Tesla's autonomous hardware suite costs ~$1k and Tesla makes its own cars Waymo's autonomous hardware suite costs > $10k and Waymo pays other companies to use their cars It's really not that complicated. Once Tesla scales, it's over

@ha1331 @JoeTegtmeyer

@ManuPr3mier @JoeTegtmeyer A banana?

Cybercab Side and Rear camera washer operation. I captured this today as Cybercabs were being loaded onto transport trucks. This is a good view of how these work!

@TukiFromKL if Musk said 2026 then we have nothing to worry about. 100% chance it wont happen

Everyone keeps saying "singularity" like they know what it means. Most of you have no idea. Let me explain it in 30 seconds and ruin your week. The singularity is the moment AI gets smart enough to improve ITSELF. No humans needed. It builds a better version of itself, that version builds an even better one, and again, and again. At a speed your brain physically cannot comprehend. The term comes from physics.. It literally means "the point where the rules stop working." Where math breaks down and where prediction becomes impossible. Ray Kurzweil said it would happen by 2045. Musk said 2026. The doomers think it ends the human race. The accelerationists think it cures cancer and solves death. But here's the thing. The real debate is whether it already started and we're just too comfortable calling it "cool new tools" to admit what we're actually looking at.

@ha1331 @JoeTegtmeyer then what's this dude holding?

@ManuPr3mier @JoeTegtmeyer How could it be, since these vehicles don't have a steering wheel?

@draslan_eth @JoeTegtmeyer huh?

@ManuPr3mier @JoeTegtmeyer No

@ha1331 @JoeTegtmeyer what's this?

@dasblitzy @EmmanuelMacron ta gueule

@EmmanuelMacron Start-up en France ? 🤔 90% de taux d'échec. 🥳

Yann Le Cun ouvre une nouvelle page de l’intelligence artificielle. C’est la France des chercheurs, des bâtisseurs et des audacieux. Bravo !

@barriere_dr 100%

Claude est vraiment meilleur que chatGPT ?