Blog BlogAnnouncements, research highlights, and perspectives on topics we find interesting.Get updatesFilter by category:companyproductresearchnewspolicytalkspodcastHow to usefully run 1,000 agents in parallelTalksApril 15, 2026Read moreBouncer: Heal your feedProductApril 8, 2026Read moreHow Offload Works: Inside the Rust CLI that sped up our tests by 6xProductApril 6, 2026Read moreA case study in testing with 100+ of Claude agents in parallelProductApril 3, 2026Read moreUsefully run hundreds of Claudes in parallel with mngrProductApril 2, 2026Read moreIntroducing Keystone: Building self-configuring agentsProductMarch 25, 2026Read moreOffload: We made agents run our test suite 6x fasterProductMarch 17, 2026Read moreVet: Prevent coding agents from making mistakesProductMarch 5, 2026Read moreBeating ARC-AGI-2 with Code EvolutionResearchFebruary 27, 2026Read moreLLM-based Evolution as a Universal OptimizerResearchFebruary 27, 2026Read moreLatchkey: Use plain old REST instead of MCPProductFebruary 27, 2026Read moreA more radical ImbueCompanyJanuary 30, 2026Read moreChoices and KnivesCompanyJanuary 17, 2026Read moreDigital Freedom Depends on Access Rights: Matt Boulos in LawfarePolicyDecember 11, 2025Read moreHow we made sandboxed coding agents 10x faster to startProductNovember 17, 2025Read moreA Healthy Ecosystem for AI Agents: Introducing the AFIPolicyNovember 7, 2025Read moreClosed vs. open AI ecosystems: Kanjun on DecoderTalksOctober 9, 2025Read moreSculptor: the missing UI for parallel coding agentsProductSeptember 26, 2025Read moreHow to build robust software systems with AI: Josh at the AI Engineer World's Fair 2025TalksJune 9, 2025Read moreBeyond the Prototype: Using AI to Write High-Quality CodeNewsJune 5, 2025Read moreEmpowering humans in the age of AICompanyMay 25, 2025Read moreSculptor: Catch and fix issues as you codeProductApril 8, 2025Read moreOne-person enterprise: Kanjun at the World Economic Forum Annual MeetingTalksMarch 4, 2025Read moreThe future of AI is human-centered: Kanjun on the Pioneers of AI podcastTalksFebruary 5, 2025Read moreGlenn McDonald: On Dungeons & Dragons, data alchemy, and radically open softwareCompanyJanuary 30, 2025Read moreKeeping humans in the loop: Josh at the Democratize Intelligence SummitTalksJanuary 24, 2025Read moreA more empowering view of general agents: Kanjun on NYSE TVTalksJanuary 23, 2025Read moreHuman agency vs. “agentic” AI: Kanjun on the NonZero podcastTalksJanuary 23, 2025Read moreThe future of AI is human-centeredNewsJanuary 5, 2025Read moreTransforming AI agents into personal collaborators: Kanjun on the NVIDIA AI PodcastTalksDecember 15, 2024Read moreThe ethical and practical future of AI: Matt Boulos at the MIT Club of Northern CaliforniaTalksNovember 14, 2024Read moreThe AI paradox: Kanjun at the 2024 Progress ConferenceTalksOctober 19, 2024Read moreRylan Schaeffer, Stanford: On investigating emergent abilities and challenging dominant research ideasPodcastSeptember 19, 2024Read moreAgency in the age of AI: Kanjun on the Minus One podcastTalksAugust 24, 2024Read more“The AI startup Imbue is building a service that will let you DIY a bot for any task, large or small, said Kanjun Qiu, its chief executive and co-founder. Rather than relying on companies to build bots for us, Qiu said Imbue's service aims to give people more control in a bot-powered world.”NewsAugust 24, 2024Read moreReinventing AI agents: Kanjun on the Gradient Dissent podcastTalksAugust 8, 2024Read moreThese Are the 10 AI Startups to Watch in 2024NewsJuly 24, 2024Read more“The lack of openness has definitely been noticeable over the past year and it's put a damper on academic investigations,” — Josh Albrecht, CTONewsJuly 24, 2024Read moreAri Morcos, DatologyAI: On leveraging data to democratize model trainingPodcastJuly 11, 2024Read moreKanjun Qiu, CEO and founder of the AI startup Imbue, which is working on agents that can reason and code, likens the state of agents to where self-driving cars were just over a decade ago.NewsJuly 5, 2024Read moreOpen-sourcing CARBS: how we used our hyperparameter optimizer to scale up to a 70B-parameter language modelResearchJune 25, 2024Read moreFrom bare metal to a 70B model: infrastructure set-up and scriptsResearchJune 25, 2024Read moreSanitized open-source datasets for natural language and code understanding: how we evaluated our 70B modelResearchJune 25, 2024Read moreTraining a 70B model from scratch: open-source tools, evaluation datasets, and learningsResearchJune 25, 2024Read moreTraining >70B LLMs on 10,000 H100 clusters: Josh on the Latent Space podcastTalksJune 25, 2024Read more“The real potential of agents, if we do a good job with invention, is that we can give individual people much more power over our computing environments." — Kanjun Qiu, CEONewsJune 18, 2024Read morePercy Liang, Stanford: On the paradigm shift and societal effects of foundation modelsPodcastMay 9, 2024Read more'We need legal protections from AI’s risks': Matt Boulos in The HillPolicyApril 21, 2024Read more“I think of it as really being able to [help us] control information environments … once we have control over information environments, we’ll feel a lot more empowered,” Qiu said. “Every single person’s vision can come to life.”NewsMarch 21, 2024Read moreSeth Lazar, Australian National University: On legitimate power, moral nuance, and the political philosophy of AIPodcastMarch 12, 2024Read more“Technology that we build reflects the values of its creators,” says Qiu. “There’s a much more human way to think about building out this technology, in a way that’s empowering to people.”NewsMarch 12, 2024Read more“What I’m most excited about is that I think [AI] is one of the biggest things that can unlock our potential.” — Kanjun Qiu, CEONewsFebruary 26, 2024Read moreTri Dao, Stanford: On FlashAttention and sparsity, quantization, and efficient inferencePodcastFebruary 8, 2024Read moreImbue joins Department of Commerce’s AI Safety Institute ConsortiumNewsFebruary 8, 2024Read moreImbue and Dell designed the custom system... to support rapid experimentation in developing AI systemsNewsNovember 28, 2023Read moreWe've partnered with Dell to build a cluster of 10,000 Nvidia H100 GPUs to train our modelsNewsNovember 28, 2023Read moreDeveloping agents that can reason and code: Kanjun and Josh on the No Priors podcastTalksNovember 16, 2023Read moreThe need for trust: Analyzing the individual comments to the NTIA’s AI RFCPolicyOctober 30, 2023Read moreMethodology for analyzing the individual comments to the NTIA’s AI RFCPolicyOctober 30, 2023Read moreImbue Raises Fresh Capital, Amazon's Alexa Fund Joins Investor LineupCompanyOctober 19, 2023Read more“We want to enable agents that people can actually use and that work reliably, and ultimately the vision is to reinvent the personal computer so that it’s actually truly personal.” Kanjun QiuNewsOctober 19, 2023Read moreWhy AI agents don’t work (yet): Kanjun on the Latent Space podcastTalksOctober 14, 2023Read moreImbue raises $200M to build AI systems that can reason and codeCompanySeptember 7, 2023Read moreImbue raises $200 Million Series BNewsSeptember 7, 2023Read moreOur technical hiring processCompanySeptember 6, 2023Read moreBryden Fogelman: On coding agents and finding meaningful work at ImbueCompanySeptember 6, 2023Read moreIn-person culture and habitsCompanySeptember 1, 2023Read moreWhere are the good policy ideas? Analyzing the NTIA's AI RFCPolicyAugust 21, 2023Read more“The real challenge is building systems with robust reasoning,” said Qiu.NewsJuly 17, 2023Read moreWhat's next in AI?: Kanjun and Reid Hoffman at Config 2023TalksJuly 6, 2023Read moreJamie Simon, UC Berkeley: On theoretical principles for how neural networks learn and generalizePodcastJune 29, 2023Read moreOn the Stepwise Nature of Self-Supervised LearningResearchJune 29, 2023Read moreScaling Laws For Every Hyperparameter Via Cost-Aware HPOResearchJune 20, 2023Read moreOur submission to the NTIA's AI accountability RFCPolicyJune 10, 2023Read moreTowards guardrails, not guidelines: a policy framework for powerful AI systemsPolicyJune 8, 2023Read moreKanjun Qiu, the CEO of Generally Intelligent, an AI research company, has been buying advanced graphic chips since last year for her own servers, allowing her to ride out the current shortage.NewsMay 29, 2023Read more“The biggest thing…that I find them to be helpful for is inspiring me as the user and helping me learn things that I wouldn’t have necessarily thought of on my own,” Josh Albrecht, CTO of Generally IntelligentNewsMay 25, 2023Read moreUpdate: The state of LLM ethical decision-makingResearchApril 1, 2023Read moreBill Thompson, UC Berkeley: On how cultural evolution shapes knowledge acquisitionPodcastMarch 29, 2023Read more“Over the next few months and years, we can expect much of the internet to get connected to large language models.” — Kanjun Qiu, CEONewsMarch 28, 2023Read moreInstead of relying exclusively on outside providers, the AI startup Generally Intelligent has begun purchasing its own graphics chips to power its research.NewsMarch 27, 2023Read moreBen Eysenbach, CMU: On designing simpler and more principled RL algorithmsPodcastMarch 22, 2023Read moreJim Fan, NVIDIA: On foundation models for embodied agents, scaling data, and why prompt engineering will become irrelevantPodcastMarch 8, 2023Read moreSergey Levine, UC Berkeley: On the bottlenecks to generalization, why simulation is doomed to succeed, and picking good research problemsPodcastMarch 1, 2023Read moreThe state of LLM ethical decision-makingResearchFebruary 14, 2023Read moreNoam Brown, FAIR: On achieving human-level performance in poker and Diplomacy, and the power of spending compute at inference timePodcastFebruary 9, 2023Read moreSugandha Sharma, MIT: On biologically inspired neural architectures, how memories can be implemented, and control theory PodcastJanuary 11, 2023Read moreNicklas Hansen, UCSD: On long-horizon planning and why algorithms don't drive research progressPodcastDecember 16, 2022Read moreJack Parker-Holder, DeepMind: On open-endedness, evolving agents and environments, online adaptation, and offline learningPodcastDecember 6, 2022Read more“But it really begs the question: what is understanding?” — Kanjun Qiu, CEONewsDecember 5, 2022Read moreCeleste Kidd, UC Berkeley: On attention and curiosity, how we form beliefs, and where certainty comes fromPodcastNovember 22, 2022Read moreArchit Sharma, Stanford: On unsupervised and autonomous reinforcement learningPodcastNovember 17, 2022Read moreChelsea Finn, Stanford: On the biggest bottlenecks in robotics and reinforcement learningPodcastNovember 3, 2022Read moreIntroducing Generally IntelligentCompanyOctober 20, 2022Read moreOur approach to safetyPolicyOctober 20, 2022Read moreAvalon: A Benchmark for RL Generalization Using Procedurally Generated Worlds (NeurIPS Spotlight Paper)ResearchOctober 20, 2022Read moreAmong those contributing to the company's $20 million in initial funding (plus over $100 million in options) is Tom Brown, former engineering lead for OpenAI's GPT-3; former OpenAI robotics lead Jonas Schneider; Dropbox co-founders Drew Houston and Arash Ferdowsi; and the Astera Institute.NewsOctober 20, 2022Read moreHattie Zhou, Mila: On supermasks, iterative learning, and fortuitous forgettingPodcastOctober 14, 2022Read moreMinqi Jiang, UCL: On environment and curriculum design for general RL agentsPodcastJuly 19, 2022Read moreOleh Rybkin, UPenn: On exploration and planning with world modelsPodcastJuly 11, 2022Read moreVICReg: Tutorial and lightweight PyTorch implementationResearchApril 21, 2022Read moreA PyTorch implementation of SIMONeResearchApril 14, 2022Read moreAndrew Lampinen, DeepMind: On symbolic behavior, mental time travel, and insights from psychologyPodcastFebruary 28, 2022Read moreYilun Du, MIT: On energy-based models, implicit functions, and modularityPodcastDecember 21, 2021Read moreMartin Arjovsky, INRIA: On benchmarks for robustness and geometric information theoryPodcastOctober 14, 2021Read moreJupyter Ascending: Sync notebooks to your editorResearchOctober 14, 2021Read moreYash Sharma, MPI-IS: On generalizability, causality, and disentanglementPodcastSeptember 24, 2021Read moreJacob Steinhardt, UC Berkeley: On machine learning safety, alignment and measurement PodcastJune 17, 2021Read moreJonathan Frankle, MIT: On the lottery ticket hypothesis and the science of deep learning PodcastJune 1, 2021Read moreVincent Sitzmann, MIT: On neural scene representations for computer vision and more general AIPodcastMay 19, 2021Read moreDylan Hadfield-Menell, UC Berkeley/MIT: On the value alignment problem in AIPodcastMay 11, 2021Read moreDrew Linsley, Brown: On inductive biases for vision and generalizationPodcastApril 1, 2021Read moreGiancarlo Kerg, Mila: On approaching deep learning from mathematical foundationsPodcastMarch 26, 2021Read moreYujia Huang, Caltech: On neuro-inspired generative modelsPodcastMarch 18, 2021Read moreA PyTorch implementation of slot attentionResearchMarch 9, 2021Read moreJulian Chibane, MPI-INF: On 3D reconstruction using implicit functionsPodcastMarch 5, 2021Read moreA new AI research lab: Josh on the Machine Learning Engineered podcastTalksFebruary 24, 2021Read moreKatja Schwarz, MPI-IS: On GANs, implicit functions, and 3D scene understandingPodcastFebruary 23, 2021Read moreJoel Lehman, OpenAI: On evolving intelligence, open-endedness, and reinforcement learningPodcastFebruary 16, 2021Read moreCinjon Resnick, NYU: On activity and scene understandingPodcastJanuary 31, 2021Read moreSarah Jane Hong, Latent Space: On neural rendering & research processPodcastJanuary 6, 2021Read moreKelvin Guu, Google AI: On language models & overlooked problemsPodcastDecember 16, 2020Read moreGenerally Intelligent #0: A podcast for deep learning researchersPodcastDecember 6, 2020Read moreAppendix for "Understanding self-supervised and contrastive learning with 'Bootstrap Your Own Latent' (BYOL)"ResearchAugust 24, 2020Read moreUnderstanding self-supervised and contrastive learning with "Bootstrap Your Own Latent" (BYOL)ResearchAugust 24, 2020Read more