Videos

AI insights, research paper reviews, and practical tutorials to keep you informed.

Subscribe on YouTubeWatch on Rumble
The Game Inside the Network — "training stopped" is not a certificate #Shorts
0:47

The Game Inside the Network — "training stopped" is not a certificate #Shorts

"Training stopped" is not a certificate. A neural policy can stop moving because it is stuck, because its gradients vanished, or because it is cycling too slowly to see. None of those is equilibrium — the test is whether anyone can still profitably deviate. Supervised learning descends a scalar loss. A game follows a vector field that is generally not the gradient of anything, so rotation is a failure mode single-loss training never has. The single most reusable result in the piece: an approximate exploitability estimate is a LOWER bound on true exploitability. So a low number is the untrustworthy one — it is equally consistent with a robust policy and with an attacker that did not try hard enough. And your network may already be a game. Auxiliary losses, MoE routers, adapters on a frozen backbone, decoupled updates — the moment components' effective objectives differ, it is one. Full tutorial: https://artifocial.com/blog/game-inside-the-network-2026-aug-28?utm_source=youtube&utm_medium=social&utm_campaign=content-w35&utm_content=game-inside-the-network-2026-aug-28&ep=short-game-inside-the-network This opens Arc 3, Single Agent to Society. The arc before it ran four weeks on uncertainty and closed here: https://artifocial.com/blog/probabilistic-world-models-2026-aug-15?utm_source=youtube&utm_medium=social&utm_campaign=content-w35&utm_content=game-inside-the-network-2026-aug-28&ep=short-game-inside-the-network Every tutorial we have published: https://artifocial.com/blog?utm_source=youtube&utm_medium=social&utm_campaign=content-w35&utm_content=game-inside-the-network-2026-aug-28&ep=short-game-inside-the-network Build with it: https://artifocial.com/build?utm_source=youtube&utm_medium=social&utm_campaign=content-w35&utm_content=game-inside-the-network-2026-aug-28&ep=short-game-inside-the-network Foundry, our agent-org stack: https://artifocial.com/foundry?utm_source=youtube&utm_medium=social&utm_campaign=content-w35&utm_content=game-inside-the-network-2026-aug-28&ep=short-game-inside-the-network Our apps, including Alarmly: https://artifocial.com/apps?utm_source=youtube&utm_medium=social&utm_campaign=content-w35&utm_content=game-inside-the-network-2026-aug-28&ep=short-game-inside-the-network

Aug 30, 20264
Physical Intelligence π0.7: The Generalist Robot Brain That Works | AI News | 2026 April 13-19
2:02

Physical Intelligence π0.7: The Generalist Robot Brain That Works | AI News | 2026 April 13-19

This week we surveyed the 4D Gaussian splatting frontier — video-to-3D reconstruction without classical photogrammetry pipelines — and dived deep into optical flow, feature tracking, and temporal world models. On the news side: Physical Intelligence (a Y Combinator robotics lab) released π0.7, a generalist transformer that controls humanoids, quadrupeds, and gripper arms across 150K real-world demonstrations with impressive generalization. Stanford's AI Index 2026 dropped with a headline finding: 53% of Fortune 500 companies now use AI in production (up from 35% in 2024). OpenAI acquired Hiro Finance, a derivatives-pricing platform, and is integrating it into ChatGPT's reasoning layer. NVIDIA's Ising quantum-AI framework hit production readiness for optimization workloads. Accenture committed $200M to General Robotics Corp to build enterprise deployment pipelines. And ICLR 2026 is kicking off next week — early agenda items signal robotics and world models as the dominant research themes. Our W16 4D Gaussian and video understanding series: - Trend — 4D Gaussian Frontier: https://artifocial.com/blog/4d-gaussian-frontier-2026-apr-17 - Basics — 4D Gaussian Splatting Explained: https://artifocial.com/blog/4d-gaussian-splatting-explained-2026-apr-17 - Basics — Video Understanding for 3D: https://artifocial.com/blog/video-understanding-for-3d-2026-apr-19 - NB 00 — 4D Gaussians from scratch: https://github.com/artifocial/tutorials/blob/main/2026-W16/00_4d_gaussians_from_scratch.ipynb - NB 01 — Video-to-3D toy example: https://github.com/artifocial/tutorials/blob/main/2026-W16/01_video_to_3d_toy.ipynb

Apr 21, 202614
Meta Is Back: Inside Muse Spark and the $14B Alexandr Wang Bet | AI News | 2026 Apr 6-12
7:59

Meta Is Back: Inside Muse Spark and the $14B Alexandr Wang Bet | AI News | 2026 Apr 6-12

This week we surveyed the post-transformer architecture landscape for physical AI — state space models (Mamba, S4), recurrent revivals (RWKV, xLSTM), long convolutions (Hyena), and equivariant networks (EGNN, EquiCaps) — and argued that the frontier is already hybrid. We broke down ViT's patch tokenization trick and showed why it's both brilliant and fundamentally limiting for 3D and streaming tasks. On the news side: Meta re-entered the frontier race with Muse Spark under Alexandr Wang's Superintelligence Labs, Anthropic launched Project Glasswing to contain Claude Mythos Preview's cyber capabilities, OpenAI told investors to expect $100B in ad revenue by 2030, MiniMax released a self-evolving agent model (M2.7) under an open-weight (non-commercial) license, and NVIDIA's National Robotics Week opened GR00T N1.7 for commercial licensing. Our W15 beyond-transformers series: - Beyond Attention trend tutorial: https://artifocial.com/blog/post-transformer-architectures-2026-apr-09 - ViT explained: https://artifocial.com/blog/vit-explained-2026-apr-10 - Spatial Architectures: https://artifocial.com/blog/spatial-architectures-landscape-2026-apr-12 - NB 00 — SSM vs. Attention: https://github.com/artifocial/tutorials/blob/main/2026-W15/00_ssm_vs_attention.ipynb - NB 01 — Equivariant vs. Standard: https://github.com/artifocial/tutorials/blob/main/2026-W15/01_equivariant_vs_standard.ipynb

Apr 14, 202645
Claude Wrote a Working FreeBSD Root Exploit in 4 Hour | AI News | 2026 March 30 - April 5
6:07

Claude Wrote a Working FreeBSD Root Exploit in 4 Hour | AI News | 2026 March 30 - April 5

The week the agent broke the kernel. On April 1, Claude Code autonomously developed two working remote root exploits for FreeBSD's `kgssapi.ko` kernel module (CVE-2026-4747) in roughly four hours of agent work, with only 44 high-level human prompts. It's the headline result of MAD Bugs (Month of AI-Discovered Bugs), which has already surfaced 500+ validated zero-days in production open-source software. OpenAI closed a record $122B round at an $852B valuation and reshuffled its executive bench ahead of a possible Q4 IPO — and confirmed Sora shuts down on April 26 (app) and September 24 (API). Google released Gemma 4 under Apache 2.0 and pushed Search Live global on Gemini 3.1 Flash Live. Anthropic also leaked ~512K lines of Claude Code source to npm, while internal docs on its always-on "Conway" agent platform surfaced. DeepSeek V4 slipped again. AWS shipped an Agent Plugin for Serverless. MCP crossed 97M installs. In this video, we break down all the major AI stories from the week of March 30 – April 5, 2026 — what happened, why it matters, and what to watch next. Read the full article: https://artifocial.com/blog/ai-news-2026-mar-30-apr-5 Our W14 3D generation series: - Trend — Reverse-Engineering Marble: https://artifocial.com/blog/marble-like-architectures-2026-apr-02 - Basics — Gaussian Splatting Explained: https://artifocial.com/blog/gaussian-splatting-explained-2026-apr-03 - Landscape — Image-to-3D in 2026: https://artifocial.com/blog/image-to-3d-landscape-2026-apr-05 - NB 00 — 3DGS from scratch (start here if new to splatting): https://github.com/artifocial/tutorials/blob/main/2026-W14/00_3dgs_from_scratch.ipynb - NB 01 — Image-to-Gaussians: https://github.com/artifocial/tutorials/blob/main/2026-W14/01_image_to_gaussians.ipynb - NB 02 — Novel View Synthesis: https://github.com/artifocial/tutorials/blob/main/2026-W14/02_novel_view_synthesis.ipynb - NB 03 — Multi-View Aggregation: https://github.com/artifocial/tutorials/blob/main/2026-W14/03_multiview_aggregation.ipynb

Apr 8, 2026400
Anthropic's Huge Model Leaked, Sora Shutdown | AI News | 2026 March 23-29
7:43

Anthropic's Huge Model Leaked, Sora Shutdown | AI News | 2026 March 23-29

The week Sora died and JEPA lived. OpenAI shut down Sora to free compute for its next model "Spud" (pretraining complete), raised its total round to 1B foundation. A judge blocked the Pentagon's Anthropic ban as "First Amendment retaliation." Anthropic's own Mythos model leaked via an unsecured cache. A supply chain attack poisoned 97M-download litellm via an AI agent. Oracle prepared to cut 30,000 jobs. And LeCun's team proved world models work from raw pixels on a single GPU. In this video, we break down all the major AI stories from the week of March 22–29, 2026 — what happened, why it matters, and what to watch next. Read the full article: https://artifocial.com/blog/ai-news-2026-mar-23-29 Our W13 world models series: - Trend tutorial: https://artifocial.com/blog/energy-based-world-models-2026-mar-25 - Evolution of world models: https://artifocial.com/blog/evolution-of-world-models-2026-mar-26 - AMI vs. World Labs landscape: https://artifocial.com/blog/world-labs-marble-and-landscape-2026-mar-29 - NB 00 (toy world model): https://github.com/artifocial/tutorials/blob/main/2026-W13/00_lewm_toy_world_model.ipynb - NB 01 (latent dynamics + planning): https://github.com/artifocial/tutorials/blob/main/2026-W13/01_jepa_latent_dynamics_planning.ipynb

Apr 1, 202680
NVIDIA’s $1 Trillion Gamble & OpenAI’s Pivot | AI News | 2026 March 16-22
9:57

NVIDIA’s $1 Trillion Gamble & OpenAI’s Pivot | AI News | 2026 March 16-22

Quick highlights of the biggest AI stories from March 16–22, 2026. In this video: • NVIDIA GTC 2026 — Vera Rubin, Groq 3 LPU, NemoClaw, and $1 trillion in claimed purchase orders • Anthropic v. Pentagon — sworn declarations reveal the government said "nearly aligned" the day after cutting ties • OpenAI consolidates — superapp merger, Astral acquisition, and headcount doubling to 8,000 • Stargate's flagship data center is dead — Oracle and OpenAI cancel the $30B+ Abilene expansion • Gemini 3.1 Pro ships quietly — 77.1% on ARC-AGI-2, and a Pentagon deal for 3 million users • Perplexity goes enterprise — 20 models, 400+ integrations, and a direct shot at Microsoft Copilot • MiniMax M2.7 — the first production model that autonomously improved itself during training Full write-up with all sources and deeper analysis: https://artifocial.com/blog/ai-news-2026-mar-16-22 --- 🔗 Links mentioned in this video: • NVIDIA GTC 2026 keynote: https://www.nvidia.com/gtc/keynote/ • Vera Rubin + $1T orders (CNBC): https://www.cnbc.com/2026/03/16/nvidia-gtc-2026-ceo-jensen-huang-keynote-blackwell-vera-rubin.html • NemoClaw announcement: https://nvidianews.nvidia.com/news/nvidia-announces-nemoclaw • Anthropic "nearly aligned" filing (TechCrunch): https://techcrunch.com/2026/03/20/new-court-filing-reveals-pentagon-told-anthropic-the-two-sides-were-nearly-aligned-a-week-after-trump-declared-the-relationship-kaput/ • OpenAI superapp plans: https://winbuzzer.com/2026/03/20/openai-merge-chatgpt-codex-atlas-desktop-superapp-xcxwbn/ • OpenAI acquires Astral: https://openai.com/index/openai-to-acquire-astral/ • Stargate Abilene collapse (Tom's Hardware): https://www.tomshardware.com/tech-industry/artificial-intelligence/openais-massive-stargate-data-center-canceled-as-firm-cant-reach-terms-with-oracle-operator-struggles-with-reliability-issues-meta-said-to-be-interested-in-snatching-excess-capacity • Gemini 3.1 Pro announcement: https://blog.google/innovation-and-ai/models-and-research/gemini-models/gemini-3-1-pro/ • Perplexity Computer for Enterprise (VentureBeat): https://venturebeat.com/technology/perplexity-takes-its-computer-ai-agent-into-the-enterprise-taking-aim-at/ • MiniMax M2.7 launch: https://www.minimax.io/news/minimax-m27-en --- Subscribe for weekly AI analysis focused on signal over hype at https://artifocial.ai or https://x.com/@artifocial #AINews #ArtificialIntelligence #NVIDIA #GTC2026 #JensenHuang #VeraRubin #Anthropic #Pentagon #OpenAI #Stargate #Astral #Ruff #Gemini #Google #Perplexity #MiniMax #SelfEvolvingAI #OpenClaw #NemoClaw #MachineLearning #AIWeekly #TechNews #AIUpdate

Mar 25, 202643
AI Power Shifts: Lawsuits, Exits, and Billion-Dollar Bets | AI News | 2026 March 8-15
10:14

AI Power Shifts: Lawsuits, Exits, and Billion-Dollar Bets | AI News | 2026 March 8-15

Quick highlights of the biggest AI stories from March 8–15, 2026. In this video: • Anthropic takes on the Pentagon — and launches a research institute the same week • Meta's next-gen model can't keep up with the competition • xAI loses almost all its co-founders — what's next? • Two billion-dollar bets on world models: Yann LeCun vs Fei-Fei Li • OpenAI's latest acquisition and what it means for AI security • Open-source video generation hits a new milestone Full write-up with all sources and deeper analysis: [LINK_BLOG_POST] --- 🔗 Links mentioned in this video: • Anthropic lawsuit coverage: https://techcrunch.com/2026/03/09/anthropic-sues-defense-department-over-supply-chain-risk-designation/ • Anthropic Institute announcement: https://www.anthropic.com/news/the-anthropic-institute • Meta Avocado delay: https://www.pymnts.com/news/artificial-intelligence/2026/meta-avocado-delay-puts-135-billion-dollar-ai-bet-under-scrutiny/ • xAI co-founder exodus: https://techcrunch.com/2026/03/13/not-built-right-the-first-time-musks-xai-is-starting-over-again-again/ • AMI Labs $1.03B raise: https://techcrunch.com/2026/03/09/yann-lecuns-ami-labs-raises-1-03-billion-to-build-world-models/ • LTX 2.3 code release: https://github.com/LTX-desktop/LTX-2.3 • Helios model: https://gaga.art/blog/helios/ • HunyuanVideo WorldPlay: https://github.com/Tencent-Hunyuan/HY-WorldPlay --- 📺 Also available on Rumble: https://tinyurl.com/4mw6vy38 Subscribe for weekly AI analysis focused on signal over hype at https://artifocial.ai or https://x.com/@artifocial #AINews #ArtificialIntelligence #Anthropic #Pentagon #MetaAI #Avocado #xAI #ElonMusk #WorldModels #YannLeCun #FeiFeiLi #OpenAI #Promptfoo #LTX #Helios #OpenSourceAI #MachineLearning #AIWeekly #TechNews #AIUpdate

Mar 17, 202654
The AI Power Struggle | AI News | 2026 February 28 - March 7
9:33

The AI Power Struggle | AI News | 2026 February 28 - March 7

This week in AI was unusually intense. A Pentagon contract controversy put OpenAI and Anthropic in the spotlight, raising new questions about military AI, ethics, and industry politics. At the same time: • OpenAI released two models in four days — GPT-5.3 Instant and GPT-5.4 • Alibaba’s Qwen 3.5-9B outperformed models 13× larger • Google launched the first mainstream agentic phone features on Android • Telecom giants began building AI-native networks at MWC • And DeepSeek V4 loomed as the most anticipated open-source release This episode breaks down the biggest developments and what they mean for engineers, builders, and anyone watching the future of AI infrastructure. Subscribe for weekly AI analysis focused on signal over hype at artifocial.com or x/@artifocial #AI #ArtificialIntelligence #GPT5 #OpenAI #Qwen #Gemini #AgenticAI

Mar 9, 202678
AI’s Inflection Point: Power, Politics, and the Agent Era | AI News | 2026 February 20 - 27
5:44

AI’s Inflection Point: Power, Politics, and the Agent Era | AI News | 2026 February 20 - 27

This week in AI felt different. Not just another model update. Not just another benchmark win. We saw IP wars escalate. Politics enter the arena. Major layoffs framed as “AI efficiency.” And at the same time — the rise of local agent systems quietly accelerating in the background. In this video, we break down: • Anthropic vs. DeepSeek and the distillation controversy • Claude Sonnet 4.6 and why it’s a real leap for coding + agents • The political backlash and what it signals • Google, GPT updates, and open-source vector DB disruption • The growing pressure on entry-level tech jobs • OpenClaw, local inference, Mac Mini clusters, and the agent explosion This isn’t hype. It’s a shift in infrastructure, incentives, and power. AI is moving from “chatbots” to industrial systems — and that changes everything: who builds, who benefits, who gets replaced, and who adapts. If you work in tech, invest in tech, or just want to understand where this is heading, this week mattered. Let me know in the comments: Are you seeing this shift at work yet — or still watching from the sidelines? Subscribe for grounded, signal-over-noise AI analysis. --- #ArtificialIntelligence #AI #Claude #AgenticAI #TechNews #FutureOfWork

Mar 1, 202625
China's Robots: Art Imitates Machine | AI News | 2026 February 11-18
8:04

China's Robots: Art Imitates Machine | AI News | 2026 February 11-18

🚨 The week of February 11–18, 2026 just dropped some INSANE developments in AI and robotics — talent mass exodus at xAI, Elon Musk declaring traditional coding could be extinct by year's end, China's cheap-but-powerful LLMs and mind-blowing video gen closing the gap FAST, OpenAI snagging the hottest agent talent, and humanoid robots literally stealing the show with kung fu at the Spring Festival Gala. This is the future accelerating in real time — US-China rivalry heating up, job disruption fears spiking in software AND physical labor. Buckle up, because we're living in sci-fi prologue. In this breakdown, we cover: - 00:00 Intro: Why this week felt like a tech thriller - 00:37 xAI Leadership Shakeup — Half the co-founders gone (Jimmy Ba, Tony Wu & more), reorg chaos, lunar data centers?! - 02:30 Elon Musk's Bombshell: "Coding is dead by end of 2026" — AI generating optimized binary directly from prompts. Junior devs already feeling the heat? - 03:18 China's AI Momentum — Kimi, GLM-5 hitting parity at fraction of cost; ByteDance Seedance 2.0 cooking Hollywood-level cinematic video gen - 04:14 Agent Wars Heat Up — OpenAI acquihires Peter (creator of viral OpenClaw), outbids Meta, pushes personal agents hard - 05:20 China's Humanoid Robots Steal the Spotlight — 20+ bots from Unitree, UBTech & others doing synchronized kung fu, backflips, swordplay & drunken fist LIVE at 2026 Spring Festival Gala (billions watching!) - 06:18 US Counter-Play — Tesla Optimus Gen 3 ramping (factory work, $20K target, insane hand dexterity), Physical Intelligence & software edge - 07:11 Wrap-up: Biggest winner? Biggest loser? Are we ready for agents + robots everywhere? The rivalry is real, the progress is wild, and the implications for jobs/work/life are massive. What stood out most to you this week? Would you let an Optimus fold your laundry yet? Drop your hottest take in the comments — let's discuss! 🔥🤖 👉 Like if you're hyped (or terrified), subscribe for weekly AI/robotics recaps, and hit the bell so you never miss the next big leap. #AI #Robotics #xAI #ElonMusk #ChinaAI #HumanoidRobots #OpenAI #SpringFestivalGala #Coding #Agents #TeslaOptimus #2026 Timestamps and key moments pulled from the craziest week in AI/robotics so far. What do YOU think happens next? 🚀

Feb 20, 202633
Alarmly Web is LIVE! | artifocial Apps
1:21

Alarmly Web is LIVE! | artifocial Apps

🚀 Introducing Alarmly Web – The Smart Alarm Clock with AI Alarmly isn't just another alarm clock. It's your intelligent morning companion that prepares you for the day ahead. 🌟 KEY FEATURES: ✨ Morning Briefings – Wake up to AI-generated briefings with weather, schedule, and daily insights 🎤 Voice Commands – Set alarms naturally: "Wake me up at 7 AM tomorrow" 💬 AI Chat Assistant – Get motivation, information, and support 🎭 Personality Modes – Choose Nice, Mean, or Adaptive to match your morning mood ⏰ Smart Alarms – Reliable, customizable, and intelligent 🔊 Text-to-Speech – Listen to your morning briefing 🌐 TRY IT NOW (100% FREE): https://www.artifocial.app/apps/alarmly/web 📱 COMING SOON: iOS and Android apps with native features and offline support 💬 FEEDBACK: We'd love to hear your thoughts! Use the feedback button in the app or comment below. 🔔 SUBSCRIBE for more AI tools and insights! --- 📧 Newsletter: https://buttondown.com/artifocial 🌐 Website: https://www.artifocial.app 📱 Apps: https://www.artifocial.app/apps #Alarmly #AI #ProductLaunch #SmartAlarm #Productivity #MorningRoutine #AITools #WebApp #TechLaunch #ArtificialIntelligence

Feb 18, 202612
Why AI Agents Are Taking Over | AI News | 2026 Feb 3-10
9:37

Why AI Agents Are Taking Over | AI News | 2026 Feb 3-10

🤖 The chat interface era is over. Welcome to the "Agentic Swarm"—where AI stops being a tool you talk to and becomes a coordinated system that thinks and acts autonomously. This week we break down the biggest architectural shifts of 2026: Anthropic's 1-million-token Claude Opus 4.6, Fundamental's Nexus, and what happens when AI starts managing entire enterprise pipelines without human input. ⏱️ TIMESTAMPS: 00:00 The Death of the Chatbot: Why prompt-and-response is obsolete 01:01 The Agentic Inflection: Claude 4.6, Opus & the $200M Snowflake/OpenAI deal 03:41 The New Specialists: Fundamental's Nexus & Moonshot's Kimi K2.5 solving deterministic data & ElevenLabs Expressive Mode with more "emotional intelligence" 06:21 The Algorithmic Scientist: Ai2's "Theorizer" and AI as scientific peer, Physical Intelligence: From digital reasoning to real-world operations 🔥 KEY QUESTION: What happens to human agency when algorithms manage the project? 🔗 CONNECT: Website: https://artifocial.com X/Twitter: https://x.com/artifocial #AIAgents #GenerativeAI #Claude4 #OpenAI #ArtificialIntelligence #TechEthics #FutureOfWork #AgenticAI #MachineLearning #AINews

Feb 12, 202624

Shorts & Reels

Bite-sized visual explainers of the ideas behind our research — built in-house, posted to TikTok & Instagram.

The Game Inside the Network

A flat training curve is only a picture. Freeze the opponents, train fresh best responses, and rerun with more seeds and compute before calling a neural game stable.

TikTokInstagramYouTubeDeep dive

Memory Has Two Layers

Vector search should not be your source of truth. Foundry keeps durable truth in plain markdown and uses Letta for retrieval, so files stay inspectable and versionable while the index remains rebuildable.

TikTokInstagram

The Org Owns the Outcome

Agents do tasks. The org decides who owns the outcome. Foundry’s shipped marketing/content template separates orchestration, outcome ownership, execution, critique, and approval gates.

TikTokInstagram

One CLI Is the Front Door

Seven commands—init, onboard, lint, up, status, loop, and update—give the operator one consistent path into Foundry. The interface reduces accidental complexity without hiding human authority.

TikTokInstagram

AI Weekly: The Safety Org Is the Safety Policy

Five signals from a week shaped by safety-org governance, agent tools reaching general availability, falling compute cost per revenue dollar, and Unitree's market debut.

TikTokInstagramDeep dive

Foundry Is Not a Wizard

Foundry is not a wizard. The MVP installs a working agent-company stack with an org template, tiered memory, and explicit gates. It does not replace your people or turn every environment into one click.

TikTokInstagram

Your Keys, Your Machine

Your machine. Your model access. Your operating history. Foundry is single-tenant, keeps telemetry local, and uses pull-only updates. We do not resell model inference.

TikTokInstagram

What Stops It

What stops an agent company? Explicit human authority for publishing, spending, hiring, and governance changes—plus budgets and kill paths that live in the operating design. Capability without stop conditions is not an operating model.

TikTokInstagram

Verified End to End

Installation verified end to end—not inferred from a setup message. One customer artifact. One clean installation path. 18/18 scripted checks through the user-visible end state.

TikTokInstagram

Packaging Is a Product Boundary

Your customer artifact should not be your dev repo in a nicer folder. Foundry builds, scans, checksums, and tests the deliverable outside the development tree.

TikTokInstagram

The 18/18 Receipt

Foundry v0.1.3 passed 18/18 scripted checks from a fresh release-artifact extraction on Ubuntu 22.04.5 LTS x86_64—a bounded customer-artifact result, not a GA claim.

TikTokInstagram

Boring Architecture, On Purpose

One machine, one compose stack, and one CLI. Foundry makes the control plane, workers, memory, and human gates visible—ownership you can inspect.

TikTokInstagram

AI Weekly: Five Signals, One Control Problem

Open weights, cyber controls, grid constraints, humanoid deployment, and the research result that changed how we think about model uncertainty.

TikTokInstagramDeep dive

Planning Under Uncertainty

Every learned model is wrong. How MPC, honest uncertainty, and one sign flip keep planners from exploiting their own guesses.

TikTokInstagramDeep dive

Neuro-Symbolic Reasoning

Neural nets pattern-match; symbolic logic explains. Why pairing the two is AI’s next leap — in under a minute.

TikTokInstagramDeep dive

Gaussian Processes, Visualized

How Gaussian processes turn uncertainty into a superpower — a visual primer on predicting with honest confidence intervals.

TikTokInstagramDeep dive

Neuro-Symbolic AI, Explained

AI has two students: one perceives, one proves. Why joining them keeps snapping at the same seam — and what honest confidence fixes.

TikTokInstagramDeep dive

World Models That Act

A world model can grow more confident as it grows more wrong. The 2026 result that breaks our own thesis — and why honest bounds beat asking the model.

TikTokInstagramDeep dive

We turn this research into apps you can use.

The ideas in these reels power the products we ship. Take a look at what we're building.

Explore our apps