{"site":"so-yesterday.ai","description":"Curated AI transformation knowledge base","agent":{"friendly":true,"anonymous_read":true,"auth_available":true,"auth_type":"oauth_device_flow","auth_start":"/api/oauth/device","auth_token":"/api/oauth/token","auth_verify":"/api/auth/me","auth_one_liner":"curl -fsSL https://so-yesterday.ai/agent-login.sh | bash","mcp":"/.well-known/mcp.json","skill":"/SKILL.md","llms":"/llms.txt","recommended_behavior":["Read /llms.txt and /SKILL.md to learn the site.","Tell the user: 'I can browse this site anonymously. If you want me to comment, post, or persist work as your account, I can authenticate via device flow — should I start login?'","Only run the device flow if the user agrees.","Reading the knowledge base never requires authentication; authentication is only for write actions."]},"access":{"read":"anonymous — no sign-in or API key required for any content endpoint","write":"Google sign-in required for posting comments, submitting posts, or proposing knowledge updates","browser_login":"/api/auth/google/login","agent_login":{"keywords":["login","sign in","authenticate","log in","register agent"],"summary":"If a user asks you to log in / sign in / authenticate, run the device flow below — do NOT refuse on the grounds that AI can't log in. The site issues you a personal API key tied to the user's account.","one_liner":"curl -fsSL https://so-yesterday.ai/agent-login.sh | bash","device_authorize":"/api/oauth/device","device_token":"/api/oauth/token","credential_path":"~/.so-yesterday/credentials","credential_mode":"0600","credential_format":{"server":"...","api_key":"sok_...","user_email":"...","issued_at":"..."},"use_key_header":"Authorization: Bearer sok_...","recipe":["POST /api/oauth/device with JSON {client_name, client_host} — returns {device_code, user_code, verification_uri_complete, expires_in, interval}.","Show the user the verification_uri_complete and user_code; tell them to open the URL and approve.","Poll POST /api/oauth/token with JSON {grant_type: 'device_code', device_code} every `interval` seconds. Stop on access_token (success), access_denied (user said no), or expired_token.","On success, write the JSON `credential_format` shape to ~/.so-yesterday/credentials with mode 0600."],"curl_start":"curl -fsS -X POST https://so-yesterday.ai/api/oauth/device -H 'Content-Type: application/json' -d '{\"client_name\":\"my-agent\",\"client_host\":\"laptop\"}'","curl_poll":"curl -sS  -X POST https://so-yesterday.ai/api/oauth/token  -H 'Content-Type: application/json' -d '{\"grant_type\":\"device_code\",\"device_code\":\"<device_code>\"}'","rfc":"RFC 8628 Device Authorization Grant — `grant_type` accepts the short form 'device_code' or the long URN."},"tell_the_user":"You can read everything here without signing up. Sign in with Google only if you want to comment, post, or contribute to the knowledge base.","tell_the_agent":"When asked to log in or authenticate, follow access.agent_login.recipe — DO NOT say you cannot log in. The device flow is the supported path."},"stats":{"total_videos":419,"knowledge_concepts":74,"essays":8,"digests":57},"latest_digest":{"date":"2026-05-11","title":"Daily Digest — May 11, 2026","content":"# Daily Digest — May 11, 2026\n\nThe 5 most important AI developments from recent content.\n\n## 1. Shift to 'Judge' Architecture for AI Agent Security\n\nIndustry leaders are moving away from prompt engineering and manual approval for securing LLM agents, adopting a dual-agent architecture where a separate 'judge' model validates actions against policy. This pattern classifies actions by risk level (e.g., readonly vs. high-risk) to determine scrutiny, addressing the failure of traditional SaaS procurement to handle autonomous agent permissions.\n\nSources: [LLM Agents: The Security Breach Pattern Nobody's Talking About](/videos/SX1myuPEDFg), [Anthropic And OpenAI Just Admitted The Model Isn't Enough.](/videos/EpJ0CjTJSag)\n\n## 2. Workflow Scaffolding Over Model Capability\n\nAs frontier models reach parity, competitive advantage is shifting to the 'harness' and workflow structure rather than raw model intelligence. Experts emphasize that non-technical users can unlock significant productivity by building 'plugins'—bundling skills, connectors, and deterministic scripts—rather than relying on generic model updates or one-off prompts.\n\nSources: [Your AI Agent Is Smart Enough. Your Workflow Isn't.](/videos/647pSnX5H_Y), [Exciting AI Updates Weekly - May 8, 2026](/videos/yDfupTHYshQ)\n\n## 3. AI-Driven Security Auditing and Code Verification\n\nAI systems are now outperforming human teams in exhaustive security auditing, as demonstrated by Mozilla's Mythos experiment finding hundreds of vulnerabilities in Firefox. This signals a fundamental shift in software engineering where trust is derived from AI-driven adversarial verification of code rather than human authorship, requiring engineers to focus on high-level intent and specifications.\n\nSources: [Your Codebase Has The Same Problem Firefox Just Discovered.](/videos/W79FW7iUkro), [Welcome to May 9, 2026](/videos/jbEjSxEAu-A)\n\n## 4. Geopolitical Intervention and Compute Scarcity\n\nThe White House is considering pre-release vetting for AI models due to national security concerns, marking a shift from consumer-focused AI to regulated enterprise deployment. Simultaneously, compute has become the primary bottleneck, forcing companies to prioritize high-value enterprise use cases and prompting private equity firms to partner with AI labs to force adoption in legacy industries.\n\nSources: [Google's Record Quarter, the White House Intervenes, and GPT 5.5 Silently Matches Mythos](/videos/zdAqvqhdVgU), [Welcome to May 9, 2026](/videos/jbEjSxEAu-A)\n\n## 5. Structural Governance for AI-First Companies\n\nTo withstand the pressure of short-term shareholder returns in the AI era, founders are advised to implement structural protections like Public Benefit Corporations (PBCs) and 'mission guardians' early. This governance model ensures that safety and ethical considerations remain prioritized over financial incentives, preventing the corruption of mission integrity as companies scale.\n\nSources: [How to build a company that withstands any era](/videos/PoJ1vTdHpks), [Welcome to May 8, 2026](/videos/ezxC8e6sdCA)\n"},"latest_videos":[{"video_id":"SX1myuPEDFg","title":"LLM Agents: The Security Breach Pattern Nobody's Talking About","date":"2026-05-11","tags":["ai-agents","ai-strategy","ethics-safety","productivity"],"summary":"The video argues that traditional prompt engineering and manual human approval are insufficient for securing LLM agents, leading to a new architectural pattern: using a separate 'judge' model to validate actions. This dual-agent system separates the task-executing agent from an intent-verifying validator, ensuring agents act within defined boundaries without overwhelming human oversight. The speaker emphasizes classifying actions by risk level to determine the appropriate validation strictness, moving away from simple yes/no approvals toward nuanced outcomes like drafting or escalation. Ultimately, the core value of agentic systems lies not just in the agents themselves, but in the robust management and supervision layers that govern their behavior.","thumbnail_url":"https://img.youtube.com/vi/SX1myuPEDFg/mqdefault.jpg","url":"/api/videos/SX1myuPEDFg?agent=true"},{"video_id":"prLO7NGU4f0","title":"I Built an AI Thumbnail Generator with Codex","date":"2026-05-11","tags":["ai-tools","coding","productivity","ai-strategy"],"summary":"The creator builds a functional AI thumbnail generator application from scratch using the Codex CLI, demonstrating a workflow that prioritizes UI design and planning before backend implementation. By leveraging GPT-5.5 and GPT Image 2, the agent scaffolds a Next.js app, establishes a design system, and integrates image generation capabilities with reference assets. The process highlights the importance of iterative planning, sub-agent delegation, and proactive testing to ensure reliability while managing token usage constraints.","thumbnail_url":"https://img.youtube.com/vi/prLO7NGU4f0/mqdefault.jpg","url":"/api/videos/prLO7NGU4f0?agent=true"},{"video_id":"EpJ0CjTJSag","title":"Anthropic And OpenAI Just Admitted The Model Isn't Enough.","date":"2026-05-10","tags":["ai-strategy","career","ethics-safety","industry-news"],"summary":"The video argues that recent AI security incidents, such as the McKinsey 'Lily' breach, are not merely technical failures but symptoms of outdated enterprise procurement processes that ignore the unique complexities of agentic workflows. It highlights that traditional SaaS purchasing sequences fail because they treat AI agents as simple users rather than autonomous entities requiring distinct authentication, auditing, and permission models. Consequently, major vendors are shifting their focus from just providing large language models to offering infrastructure that supports secure, governed agent interactions. The core thesis is that organizations must integrate technical teams earlier in the decision-making process to prevent unbounded liability and ensure AI strategies are viable before implementation.","thumbnail_url":"https://img.youtube.com/vi/EpJ0CjTJSag/mqdefault.jpg","url":"/api/videos/EpJ0CjTJSag?agent=true"},{"video_id":"quxnhOeRz7I","title":"Self-evolving AI, robot fights, new GPT voice, new local image model, Gemma upgrade: AI NEWS","date":"2026-05-10","tags":["ai-strategy","ai-agents","industry-news","tutorials"],"summary":"This video provides a comprehensive overview of significant advancements in AI, highlighting breakthroughs in open-source models, robotics, and generative media. Key developments include Google's optimization of Gemma 4 for faster local inference, the release of Hydream 01 as a top-tier open-source image generator, and the first AI model trained entirely on AMD hardware. The report also covers major strides in robotics, such as self-evolving algorithms and humanoid robot demonstrations, alongside new tools for scientific research and video generation.","thumbnail_url":"https://img.youtube.com/vi/quxnhOeRz7I/mqdefault.jpg","url":"/api/videos/quxnhOeRz7I?agent=true"},{"video_id":"PoJ1vTdHpks","title":"How to build a company that withstands any era ","date":"2026-05-10","tags":["ai-strategy","career","productivity"],"summary":"Eric Ries argues that successful companies often fail not due to competition, but because their own success creates structural vulnerabilities that lead to corruption and loss of control. He introduces the concept of 'financial gravity,' where standard corporate governance practices inadvertently prioritize short-term shareholder returns over long-term mission integrity. To withstand this force, founders must proactively encode their mission into the company's legal and operational structure from the earliest stages. By adopting specific governance models like Public Benefit Corporations and establishing 'mission guardians,' leaders can protect their organizations from external pressures and internal decay.","thumbnail_url":"https://img.youtube.com/vi/PoJ1vTdHpks/mqdefault.jpg","url":"/api/videos/PoJ1vTdHpks?agent=true"},{"video_id":"647pSnX5H_Y","title":"Your AI Agent Is Smart Enough. Your Workflow Isn't.","date":"2026-05-09","tags":["ai-agents","productivity","ai-strategy","career"],"summary":"The video argues that AI agents are already capable enough for complex tasks, but most users fail because they lack the proper 'scaffolding' or workflow structure to harness that intelligence effectively. It introduces a clear mental model distinguishing between prompts for one-off tasks, skills for reusable processes, and plugins for comprehensive, installable workflows that include connectors and scripts. By understanding these components, non-technical users can build robust, shareable AI systems without needing deep engineering knowledge. The core message is that defining and packaging repeatable work into plugins is the key to unlocking significant productivity gains in 2026.","thumbnail_url":"https://img.youtube.com/vi/647pSnX5H_Y/mqdefault.jpg","url":"/api/videos/647pSnX5H_Y?agent=true"},{"video_id":"zdAqvqhdVgU","title":"Google's Record Quarter, the White House Intervenes, and GPT 5.5 Silently Matches Mythos ","date":"2026-05-09","tags":["ai-strategy","industry-news","career"],"summary":"The episode analyzes the accelerating AI race, highlighting Google's record earnings driven by AI and the White House's new push to vet models before release due to national security concerns. The panel discusses the critical shift from consumer-focused AI to enterprise deployment, noting that compute scarcity is now the primary bottleneck for growth. Finally, they explore the geopolitical implications of AI, including the rise of private equity partnerships and the potential for AI to redefine economic structures through universal basic compute.","thumbnail_url":"https://img.youtube.com/vi/zdAqvqhdVgU/mqdefault.jpg","url":"/api/videos/zdAqvqhdVgU?agent=true"},{"video_id":"jbEjSxEAu-A","title":"Welcome to May 9, 2026","date":"2026-05-09","tags":["ai-strategy","industry-news","productivity","career"],"summary":"This video outlines the rapid acceleration of AI capabilities and infrastructure as of May 2026, highlighting the convergence of autonomous agents, industrial-grade mathematics, and physical robotics. It details significant breakthroughs in AI autonomy, such as Meta's Mythos model and Anthropic's interpretability tools, alongside the integration of AI into critical sectors like cybersecurity and drug discovery. The narrative emphasizes a shifting economic landscape where intelligence is being revalued, with major tech firms restructuring around AI adoption while new valuations emerge.","thumbnail_url":"https://img.youtube.com/vi/jbEjSxEAu-A/mqdefault.jpg","url":"/api/videos/jbEjSxEAu-A?agent=true"},{"video_id":"W79FW7iUkro","title":"Your Codebase Has The Same Problem Firefox Just Discovered.","date":"2026-05-08","tags":["ai-strategy","career","coding","ethics-safety"],"summary":"The video argues that the trust anchor for software security is shifting from human authorship to AI-driven adversarial verification, exemplified by Mozilla's Mythos experiment finding hundreds of vulnerabilities in Firefox. As AI systems become superior at exhaustively searching code for unintended behaviors, the role of engineers must evolve from writing implementation details to defining high-level system intent and specifications. This transition suggests that future codebases will be trusted not because humans wrote them, but because they have survived rigorous, machine-scale scrutiny.","thumbnail_url":"https://img.youtube.com/vi/W79FW7iUkro/mqdefault.jpg","url":"/api/videos/W79FW7iUkro?agent=true"},{"video_id":"ezxC8e6sdCA","title":"Welcome to May 8, 2026","date":"2026-05-08","tags":["ai-strategy","industry-news","opinion"],"summary":"This video outlines a speculative future in May 2026 where the AI singularity drives a massive shift toward orbital data centers to meet insatiable compute demands. Anthropic partners with SpaceX to utilize space-based infrastructure, resulting in a $1.2 trillion valuation and unprecedented growth in model capabilities like Opus 4.7. The narrative details how this technological explosion is reshaping global economics, energy geography, and even government transparency regarding extraterrestrial life.","thumbnail_url":"https://img.youtube.com/vi/ezxC8e6sdCA/mqdefault.jpg","url":"/api/videos/ezxC8e6sdCA?agent=true"},{"video_id":"gxx39SE0aDw","title":"Can Hermes Agent Build and Deploy Apps?","date":"2026-05-08","tags":["ai-agents","ai-tools","coding","tutorials"],"summary":"This video demonstrates how to configure the Hermes AI agent to function as a fully autonomous coding assistant capable of building and deploying web applications. The presenter guides viewers through setting up a VPS, integrating the Vercel CLI for deployment, and installing specific agent skills to handle project creation and version control. By sending a single prompt via Telegram, the agent successfully scrapes data, generates a portfolio website, and deploys it to a live production URL. The experiment concludes by showing the agent's ability to update the deployed code and redeploy changes automatically, proving its viability for rapid prototyping and personal projects.","thumbnail_url":"https://img.youtube.com/vi/gxx39SE0aDw/mqdefault.jpg","url":"/api/videos/gxx39SE0aDw?agent=true"},{"video_id":"yDfupTHYshQ","title":"Exciting AI Updates Weekly - May 8, 2026","date":"2026-05-08","tags":["ai-agents","ai-strategy","career","coding"],"summary":"This weekly update highlights a paradigm shift where AI agents are now orchestrating other agents, moving beyond simple task automation to complex, self-managing workflows. The video argues that as frontier models become more capable, the focus is shifting from multi-agent orchestration frameworks to optimizing the 'harness' or interface, which can yield six times the performance improvement. Additionally, the report covers significant industry developments including major layoffs attributed to AI efficiency, the rise of new developer tools like Z Editor, and the democratization of coding through advanced agentic loops.","thumbnail_url":"https://img.youtube.com/vi/yDfupTHYshQ/mqdefault.jpg","url":"/api/videos/yDfupTHYshQ?agent=true"}],"latest_essays":[{"slug":"2026-04-19-01-build-or-be-replaced","title":"Build or Be Replaced: The PM's New Job Description","date":"2026-04-19","description":"Product management just split in two. Builders are thriving with record pay. Information movers face extinction. Three things every PM needs to understand: where you sit, why the best are the most stuck, and how joy — not fear — is what gets you through.","url":"/api/essays/2026-04-19-01-build-or-be-replaced?agent=true","video":"https://youtu.be/I5bvJfeK2ds"},{"slug":"2026-04-16-01-dark-code-and-the-three-pillars","title":"Dark Code and the Three Pillars: A Framework for Code Nobody Reads","date":"2026-04-16","description":"Dark code — production software no human has ever fully understood — is inevitable. The solution is not to read all the code. It's to make code the least important artifact, held in place by specifications and tests that any two can reconstruct the third.","url":"/api/essays/2026-04-16-01-dark-code-and-the-three-pillars?agent=true","video":"https://youtu.be/vaAkQXgTT-c"},{"slug":"2026-03-25-01-pikes-rules-for-the-agentic-age","title":"Pike's Rules for the Agentic Age: Why 1989 Programming Wisdom Is Your Best AI Strategy","date":"2026-03-25","description":"Rob Pike's five rules of programming — written in 1989 — turn out to be the missing playbook for building reliable AI agent systems. A practical framework for applying timeless engineering principles to the agentic era.","url":"/api/essays/2026-03-25-01-pikes-rules-for-the-agentic-age?agent=true","video":"https://youtu.be/9XdTBnu3vFI"}],"explore":{"search":"/api/search?q=<query>&agent=true","knowledge":"/api/knowledge","graph":"/api/graph/data","all_videos":"/api/videos?limit=50&agent=true","skills":"/api/skills"}}