⚜️✨ QUEBEC.IA — Frontière. IA‑First. Souveraine. Le Québec entre dans l’ère IA‑First. QUEBEC.IA avance l’IA de frontière, l’infrastructure souveraine, les agents autonomes, la sécurité, l’assurance et la gouvernance stratégique. quebec.ai

GoalOS AGIALPHA Ascension is an experimental framework for a persistent, goal-seeking, self-improving intelligence system that accumulates capabilities, evidence, and economic value over time. GitHub: github.com/MontrealAI/goalos… #AGIALPHA

In awe of SpaceX and its story - past, present and the future. You can think about it in 10 different ways and continue re-blowing your mind in circles. Huge congrats to the team! 🚀

GoalOS-native α‑AGI Ascension using AGIALPHA GitHub : github.com/MontrealAI/goalos… #AGIALPHA #AGIAscension

Quite interesting thread on capabilities of real biological neurons (spoiler: they're way more capable than classical artificial neurons in a perceptron) . Nice work @IdoAizenbud and collaborators!

L’AGI n’est peut-être pas la ligne d’arrivée. C’est peut-être le coup de départ. Google DeepMind vient de publier un rapport majeur : From AGI to ASI La question centrale n’est pas seulement : l’AGI de niveau humain transformerait-elle la société ? La vraie question est : que se passe-t-il après ? Si l’IA atteint une intelligence générale comparable à celle d’un humain, le progrès ralentit-il, plafonne-t-il, ou continue-t-il vers l’ASI — une superintelligence artificielle plus capable que de grandes organisations humaines coordonnées ? Le rapport cartographie quatre trajectoires possibles. 1. Scaler l’AGI Plus de compute. Plus d’instances. Plus de vitesse. Plus de recherche au test-time. Plus de déploiement parallèle. Même si les modèles individuels plafonnent, exécuter des millions de systèmes de niveau AGI, plus vite et moins cher, pourrait créer un nouveau régime de capacités. 2. Changer de paradigme Nouvelles architectures, nouveaux objectifs d’entraînement, meilleures mémoires, nouveaux agents, nouveaux environnements, nouveaux outils. Une rupture de paradigme pourrait produire des sauts de capacité que les courbes actuelles ne prédisent pas. 3. Amélioration récursive L’IA pourrait accélérer la recherche en IA elle-même : meilleures données meilleures architectures meilleures évaluations meilleurs environnements RL meilleurs agents Si l’IA améliore le processus qui améliore l’IA, la dynamique change. 4. Collectifs multi-agents à grande échelle L’ASI ne ressemblera peut-être pas à une seule intelligence monolithique. Elle pourrait émerger de vastes sociétés de systèmes IA : agents spécialisés, planificateurs, chercheurs, vérificateurs, utilisateurs d’outils, institutions artificielles coordonnées à vitesse machine. Ce point est crucial. La superintelligence pourrait être moins un oracle isolé qu’une organisation artificielle plus cognitivement capable que n’importe quelle organisation humaine. Le rapport est prudent. Il ne prétend pas qu’une explosion d’intelligence est garantie. Il ne prétend pas que l’ASI est inévitable. Il étudie des trajectoires, des frictions, des goulots d’étranglement et des questions de recherche ouvertes. Mais l’implication stratégique est claire : il ne faut pas traiter l’AGI comme un événement unique. Elle pourrait être une transition de phase dans un continuum plus long d’intelligence machine. Le monde ne vivra peut-être pas “un grand choc”. Il pourrait vivre une série de transformations provoquées par l’IA dans la science, la technologie, l’économie, la gouvernance et la culture. Ce n’est pas le même problème de préparation. Crédit complet aux auteurs : Tim Genewein, Matija Franklin, Alexander Lerchner, Laurent Orseau, Samuel Albanie, Adam Bales, Cole Wyeth, Stephanie Chan, Iason Gabriel, Joel Z. Leibo, Allan Dafoe, Marcus Hutter, Thore Graepel, Shane Legg. Rapport : From AGI to ASI arxiv.org/abs/2606.12683 J’attache la première page, car le cadrage mérite d’être lu directement. La prochaine frontière n’est pas seulement de construire l’AGI. C’est de comprendre ce que l’AGI rend possible ensuite. #AGI #ASI #IntelligenceArtificielle #AIResearch #GoogleDeepMind #Superintelligence

GoalOS AGIALPHA Ascension is an experimental framework for a persistent, goal-seeking, self-improving intelligence system that accumulates capabilities, evidence, and economic value over time. GitHub : github.com/MontrealAI/goalos… #AGIALPHA

RSI was missing a constitution. Not “AI improves itself.” Proof-gated evolution: Aim → Act → Prove → Evolve. No proof, no evolution. No eval, no propagation. No rollback, no release. AEP-001: GoalOS github.com/MontrealAI/proof-… #GoalOS #MontrealAI

Artificial intelligence is not replacing human intuition in maths and physics, but reimagining how questions are asked, explored and understood go.nature.com/4vG4Vsl

If you’re not inside, you’re outside. #AGIClub

🔱 If you’re not inside, you’re outside ✨ #AGIClub #AIAgents #ENS

AI 101 Masterclass 📖✨ By: QUEBEC.AI Website: quebec.ai #AI101 #MontrealAI #QuebecAI

[ The Businesses Of AGI ] AGI Agents 👾 are potential titans of global wealth. “To capture 10% of the $14 quadrillion AGI market, approximately 140,000 businesses of AGI would be needed.” By QUEBEC.AI: quebecartificialintelligence… #AGIFirst #MontrealAI #QuebecAI

Concentration of power, capabilities and economic wealth is the biggest risk in AI. We need open science and open-source more than ever!

The AI infrastructure boom is generating strong demand for skilled blue-collar workers. In fact, there’s a shortage of electricians, fiber technicians, and mechanical tradespeople needed to build and maintain AI data centers. Meta’s new $115M America’s Workforce Academy provides paid training plus job guarantees for exactly these roles. This is the kind of practical jobs training program that we need more of.

I’m incredibly excited that Fable is now available for everyone! I’ve been blown away by how smart it is - it one-shots entire PRs for me, finds obscure bugs and has written all my code since I started using it.

The Transformer scaled intelligence inside models. The next frontier may be scaling intelligence across organizations. I’m sharing my paper: AGI ALPHA: A Scalable Substrate for Intelligence Organizations The core thesis: AI progress will not be defined only by stronger models. It will be defined by systems that convert model capability into verified work, verified work into reusable capability, and reusable capability into productive capacity. AGI ALPHA proposes an organizational substrate above models: agents jobs validators tools memory markets settlement governance capacity allocation The point is not “more agents talking.” The point is proof-bearing machine labor. A job should be bounded. A tool action should be traced. A result should be validated. A capability should be reusable. A settlement should be auditable. A claim should require evidence. In this framing, intelligence is not merely benchmark performance. It is governed, evidence-producing, compounding institutional work. The paper’s central distinction: Transformer = scalable substrate for intelligence inside models AGI ALPHA = scalable substrate for intelligence across governed organizations This is also why the paper is claim-bounded. It does not claim achieved AGI, ASI, empirical SOTA, autonomous sovereignty, energy abundance, or civilization-scale capability. It proposes a testable architecture and evaluation program. The burden of proof is explicit: real tasks baselines ProofBundles replay logs validator reports cost and safety ledgers delayed outcomes independent reproduction No Evidence Docket, no strong empirical claim. That discipline matters. Because the future of agentic AI will not be won by systems that merely appear autonomous. It will be won by systems that can prove what they did, what it cost, what risks were controlled, what was learned, and whether the resulting capability compounds. AGI ALPHA is my attempt to formalize that substrate. A validator-gated, proof-bearing, RSI-governed architecture for turning machine intelligence into auditable machine labor — and eventually into reusable capability, infrastructure, science, compute, and useful energy capacity. Full credit: Vincent Boucher President, MONTREAL.AI and QUEBEC.AI Paper: AGI ALPHA: A Scalable Substrate for Intelligence Organizations github.com/MontrealAI/agialp… I’m attaching the first page because the thesis is worth reading directly. The next frontier is not just bigger models. It is institutions that make intelligence verifiable. #AGIAlpha #QuebecAI #SovereignAI #ArtificialIntelligence #AIResearch #Agents

Introducing Claude Fable 5: a Mythos-class model that we’ve made safe for general use. Its capabilities exceed those of any model we’ve ever made generally available.

We've known about LLM test-time compute scaling since @OpenAI o1. Yet 2 years later labs still report scalar evals for models; safety orgs are still surprised when a scaffold does better via 100x inference; and RSPs still ignore inference budget when deciding critical thresholds.

RSI was missing a constitution. Not “AI improves itself.” Proof-gated evolution: Aim → Act → Prove → Evolve. No proof, no evolution. No eval, no propagation. No rollback, no release. AEP-001: GoalOS github.com/MontrealAI/proof-… #GoalAI #MontrealAI #QuebecAI