Marc Mathieu

Happy New Year, and welcome to 2026! As we set intentions for the year ahead, maybe this is the year we ought to seriously think about reclaiming our right brain! Dr. Iain McGilchrist's groundbreaking work on the divided brain isn't just theory—it's a warning. He argues our modern world, with its relentless focus on data, efficiency, and instant gratification, is subtly but surely privileging the Left Hemisphere. Ready to make reclaiming our Right Hemisphere a 2026 Priority?

In era of pretraining, what mattered was internet text. You'd primarily want a large, diverse, high quality collection of internet documents to learn from. In era of supervised finetuning, it was conversations. Contract workers are hired to create answers for questions, a bit like what you'd see on Stack Overflow / Quora, or etc., but geared towards LLM use cases. Neither of the two above are going away (imo), but in this era of reinforcement learning, it is now environments. Unlike the above, they give the LLM an opportunity to actually interact - take actions, see outcomes, etc. This means you can hope to do a lot better than statistical expert imitation. And they can be used both for model training and evaluation. But just like before, the core problem now is needing a large, diverse, high quality set of environments, as exercises for the LLM to practice against. In some ways, I'm reminded of OpenAI's very first project (gym), which was exactly a framework hoping to build a large collection of environments in the same schema, but this was way before LLMs. So the environments were simple academic control tasks of the time, like cartpole, ATARI, etc. The @PrimeIntellect environments hub (and the `verifiers` repo on GitHub) builds the modernized version specifically targeting LLMs, and it's a great effort/idea. I pitched that someone build something like it earlier this year: x.com/karpathy/statu… Environments have the property that once the skeleton of the framework is in place, in principle the community / industry can parallelize across many different domains, which is exciting. Final thought - personally and long-term, I am bullish on environments and agentic interactions but I am bearish on reinforcement learning specifically. I think that reward functions are super sus, and I think humans don't use RL to learn (maybe they do for some motor tasks etc, but not intellectual problem solving tasks). Humans use different learning paradigms that are significantly more powerful and sample efficient and that haven't been properly invented and scaled yet, though early sketches and ideas exist (as just one example, the idea of "system prompt learning", moving the update to tokens/contexts not weights and optionally distilling to weights as a separate process a bit like sleep does).

Navigate the Agentic Frontier with Emergence Agents Creating Agents™ x.com/i/broadcasts/1…

In 5 years, Amazon has moved from 1 robot for 6 human employees to now… 1 robot for 2 human employees, ie >750,000 ‘coworker machines’. Equipping them with a sense of touch is a game changer… humanoids next?!

Fantastic panel with @audreyt @Tsihanouskaya @oliverroepke interviewed by @AlexKouchner at the #AIActionSummit2025 on AI&Democracy! Thank you @Make_org_Global @AliciaCombaz @axeldauchez for bringing the topic of democracy aka #humanity to the summit! AI can - and must - be designed to be indeed pro social pro society pro democracy … in summary… pro humanity! And because that won’t be the responsibility of the few, it falls in the hand of the many! Sign the Ai&Democracy Manifesto and join the Global Alliance… NOW!

🔥 NVIDIA just shook the AI world at #CES2025! NVIDIA CEO Jensen Huang JUST announced jaw-dropping breakthroughs. Here are the top 15 key highlights you can’t afford to miss: (wait till you see #15): 👇

Digital Workers are The Next Big Thing. And it’s not because they’ll be digital that they won’t be real. Real workers working ‘hand in hand’ with us - in our professional AND personal lives! See you at Dreamforce!!!

Next-gen AI apps are just a click away for every Salesforce Trailblazer. Craft AI that transforms enterprises with Salesforce! 🚀 Mix Copilot Builder, Model Builder, & Prompt Builder to revolutionize workspaces, supercharge employees, & empower customers.#AIRevolution #SalesforceMagic ✨🛠️

Unleash AI across your enterprise with a whisper. 🌪️ Build models effortlessly with Salesforce Model Builder & turn data into action with Salesforce Data Cloud. From Snowflake to Red Shift to Big Query to Databricks to Zero Copy, it’s enterprise intelligence, revolutionized. 🧠💥 #AIUnleashed #SalesforceModelBuilder

Need towels at your customer's cabana? Just say the word. 🌟 Einstein Copilot Builder is redefining 'hands-free' in enterprise software, transforming your data and workflow into action. 🗣️✨ Talk to it, from desktop to mobile, and witness enterprise magic. #EinsteinCopilot #EnterpriseRevolution

Imagine a world where your business insights just... appear. ✨ Salesforce Copilot does just that—integrating all your sales, service, & marketing data across platforms like Snowflake, Databricks, & more—straight through Salesforce Data Cloud. Zero Copy, full insights. See magic happen! #SalesforceMagic #DataCloudRevolution 🚀📊

In the era of AI, @Salesforce stands strong with our 'Trusted AI' principles. Discover how we're building an ethical, transparent, and sustainable future. #ResponsibleAI politi.co/3Qn9Fkt

My favorite part of this @Benioff Tweet - apart from the great Einstein1 demo - is @elonmusk “interesting’ comment. Indeed @elonmusk: Einstein1 rocks 🚀😉 !

It was great to spend 10 days with @SalesforceJapan delivering our #GenerativeAI keynote at World Tour Tokyo as well as meeting numerous customers and government entities interested in transforming for AI. This is the new Success formula: a Trusted AI + Data + CRM!!!

Great to be part of @salesforce AI Day in NY this week. Thanks @Benioff @clarashih @Stallapr23 @thatstokes @baxterkb for leading the charge of ‘the new Salesforce’! Looking forward to helping all Industry TrAIlblazers transform into Trust-at-the-Core, AI-First Companies!

This time it’s different

🗓️ Mark your calendars for #SalesforceAI Day, Trailblazers! Tune in on June 12 at 1:30 p.m. ET to be inspired by the future of trusted enterprise AI and what it can do for you. ⏰ Set a reminder to stream the future. 👇twitter.com/i/events/16658…

An AI will not replace a human being… but an AI that helps a human being walk again… that’s pretty cool!

Super thrilled about @salesforce's latest game-changers: #SlackGPT, #EinsteinGPT, #ServiceGPT, and #TableauGPT! 🤩 The integration of generative AI is taking customer connections to a whole new level. With Einstein's trillion predictive transactions a week, salesforce is set for remarkable generative experiences! ❤️🤖

Super thrilled about @salesforce's latest game-changers: #SlackGPT, #EinsteinGPT, #ServiceGPT, and #TableauGPT! 🤩 The integration of generative AI is taking customer connections to a whole new level. With Einstein's trillion transactions a week, salesforce is set for remarkable experiences! ❤️🤖