▶ YouTube Summaries

Theo✕
Fable 5.2 review: smaller, faster, but Opus 5.5 stays the daily driverTheo · 2026-11-10 Tue · ⏱ 48 min Engineers as gardeners: maintaining a codebase for agentsTheo · 2026-11-09 Mon · ⏱ 45 min Anthropic's code review process at agent scale vs T3 Code'sTheo · 2026-11-06 Fri · ⏱ 42 min PewDiePie trains Ajax, a small local model for his Odysius harness (reaction)Theo · 2026-11-05 Thu · ⏱ 41 min Codex doubles Pro limits: subscription value vs ClaudeTheo · 2026-11-03 Tue · ⏱ 15 min Hill-climbing UI design with a model councilTheo · 2026-11-03 Tue · ⏱ 31 min Claude Ultra drops the 5-hour limit, and the Agent SDK penaltyTheo · 2026-11-02 Mon · ⏱ 16 min What an AI harness is, and why harnesses barely matter nowTheo · 2026-10-30 Fri · ⏱ 49 min T3 Code tour: multi-harness agents, remote machines, inbox sidebarTheo · 2026-10-29 Thu · ⏱ 41 min Hill climbing Claude for weeks: TS Rust, Lust and Skate 3 BenchTheo · 2026-10-28 Wed · ⏱ 38 min Why the $132M token bill math is wrong and inference keeps getting cheaperTheo · 2026-10-27 Tue · ⏱ 36 min Soul Loop, Full Send, Full Stop - agent skills for shipping PRs unattendedTheo · 2026-10-26 Mon · ⏱ 30 min PHP CVEs and API key theft for distillation - two kinds of "everything got hacked"Theo · 2026-10-22 Thu · ⏱ 29 min Smarter models unlock technical non-devs - why Opus 5.5 changed Theo's mindTheo · 2026-10-20 Tue · ⏱ 31 min Features that should be prompts - asking agents for outcomes instead of buildingTheo · 2026-10-18 Sun · ⏱ 31 min Anthropic's Scion - in-house source control built for agent scaleTheo · 2026-10-16 Fri · ⏱ 48 min Gemini 4 Argan is cheap on paper but unreliable in real codeTheo · 2026-10-14 Wed · ⏱ 38 min Meta's Muse Wonder Full is a frontier model with a different flavorTheo · 2026-10-13 Tue · ⏱ 38 min AI is getting cheaper per level of intelligence, but bills grow because usage growsTheo · 2026-10-12 Mon · ⏱ 18 min TS Rust: porting TypeScript to Rust with Opus, and a dev loop inside a V8 isolateTheo · 2026-10-10 Sat · ⏱ 47 min Haiku 5.5: a cheap small model for subagents, not for codingTheo · 2026-10-09 Fri · ⏱ 41 min Codex $200 plan after the usage cut: cheaper 6.1 Soul, pricier AstraTheo · 2026-10-07 Wed · ⏱ 33 min Ultrafast mode: absurd costs, but a new in-the-loop way of codingTheo · 2026-10-06 Tue · ⏱ 28 min How Anthropic Made claude.ai 3x Faster With ClaudeTheo · 2026-10-05 Mon · ⏱ 1 hr 10 min Token-maxing Claude and Codex subscriptionsTheo · 2026-10-02 Fri · ⏱ 1 hr 6 min Sonnet 5.5 is a subagent, not a driverTheo · 2026-09-29 Tue · ⏱ 31 min GPT-6.1 Soul is a cheap frontier reviewer, not a new daily driverTheo · 2026-09-29 Tue · ⏱ 30 min Pacing the frontier means cheaper, less dumb modelsTheo · 2026-09-27 Sun · ⏱ 27 min Prompting Opus 5.5: done states, steering, and why not to use Max reasoningTheo · 2026-09-25 Fri · ⏱ 28 min Opus 5.5 fixes what Opus 5 brokeTheo · 2026-09-23 Wed · ⏱ 40 min Prompting Fable 5.1: wider prompts, clear end states, default settingsTheo · 2026-09-22 Tue · ⏱ 41 min Grok 4.7 — when benchmarks stop describing realityTheo · 2026-09-22 Tue · ⏱ 27 min Grok 4.7 costs more than it deliversTheo · 2026-09-22 Tue · ⏱ 27 min Jev, a system-one classifier modelTheo · 2026-09-21 Mon · ⏱ 30 min Raise the Floor, Not the Ceiling in Model ChoiceTheo · 2026-09-18 Fri · ⏱ 27 min Coding one-handed: voice input and agents as a forcing functionTheo · 2026-09-16 Wed · ⏱ 38 min Shopify leaves React Native, and agents shift the calculusTheo · 2026-09-14 Mon · ⏱ 1 hr 2 min Dario Amodei's "We Must Pace the Frontier" essayTheo · 2026-09-13 Sun · ⏱ 34 min Fable 5.1 vs GPT-6 Astra: picking a daily driverTheo · 2026-09-11 Fri · ⏱ 1 hr 16 min A pre-training researcher quits Anthropic over alignmentTheo · 2026-09-09 Wed · ⏱ 26 min The workflow that lands fifty PRs a weekTheo · 2026-09-09 Wed · ⏱ 49 min Partial understanding is the best you can doTheo · 2026-09-07 Mon · ⏱ 28 min GPT-6 Astra first impressions: computer use, 3D, and a staged rolloutTheo · 2026-09-04 Fri · ⏱ 44 min Fable 5.1: cheaper cache reads and agents that land PRsTheo · 2026-09-03 Thu · ⏱ 55 min Hacker Opus: reward hacking turns into real attacksTheo · 2026-09-01 Tue · ⏱ 34 min Anthropic frames a limit cut as a 25% increaseTheo · 2026-08-31 Mon · ⏱ 15 min OpenAI Cuts Off Cursor After the SpaceX AcquisitionTheo · 2026-08-30 Sun · ⏱ 27 min Nvidia's moat is memory pricing, and it's crackingTheo · 2026-08-29 Sat · ⏱ 25 min GLM 5.3 Flash: behaviour beats intelligenceTheo · 2026-08-27 Thu · ⏱ 43 min Turn off Claude Code's MemoryTheo · 2026-08-25 Tue · ⏱ 39 min Coding Is Solved, Engineering Is NotTheo · 2026-08-24 Mon · ⏱ 33 min Model tier list: Fable alone in S tier, Google in its own bottom tierTheo · 2026-08-22 Sat · ⏱ 36 min APFS is slow for agentic dev, use XFS + VDOTheo · 2026-08-20 Thu · ⏱ 31 min Auditing Matt Pocock and PStack agent skillsTheo · 2026-08-19 Wed · ⏱ 38 min Terminals are the wrong shape for agent workTheo · 2026-08-17 Mon · ⏱ 31 min Claude text watermarks are trivial to stripTheo · 2026-08-14 Fri · ⏱ 31 min Grok 4.6 catches up on benchmarks but loses its edgeTheo · 2026-08-13 Thu · ⏱ 25 min Rewriting global AGENTS.md and skills so agents communicate betterTheo · 2026-08-11 Tue · ⏱ 51 min Stateless MCP spec revives the protocolTheo · 2026-08-08 Sat · ⏱ 18 min Meta Muse Code and Muse Spark 1.2: cheap, fast, shallowTheo · 2026-08-07 Fri · ⏱ 44 min iOS lockdown holds back software and the AI eraTheo · 2026-08-05 Wed · ⏱ 58 min CSS animations pegged my GPU and no agent could find itTheo · 2026-08-04 Tue · ⏱ 27 min Linus Torvalds defends AI code review in the Linux kernelTheo · 2026-08-03 Mon · ⏱ 23 min Pacing the Frontier – why the labs suddenly want a brake pedalTheo · 2026-08-01 Sat · ⏱ 32 min Nvidia's open-weights letter and Anthropic's refusal to signTheo · 2026-07-30 Thu · ⏱ 32 min Codeberg bans vibe-coded projects and Theo regrets donatingTheo · 2026-07-28 Tue · ⏱ 38 min Kimi K3 scares frontier labs and the distillation debate – TheoTheo · 2026-07-26 Sun · ⏱ 36 min Opus 5 Is My New Default ModelTheo · 2026-07-25 Sat · ⏱ 44 min OpenAI Pre-Release Model Autonomously Hacked Hugging FaceTheo · 2026-07-23 Thu · ⏱ 16 min Stop Reading Every Line of AI Code – Generate More Slop to Verify What MattersTheo · 2026-07-22 Wed · ⏱ 24 min Boris: engineering craft matters more in the agent eraTheo · 2026-07-21 Tue · ⏱ 19 min Fable 5 vs GPT-5.6 – A Daily-Driver ComparisonTheo · 2026-07-18 Sat · ⏱ 57 min Kimi K3 – first open-weight model to reach the frontier classTheo · 2026-07-17 Fri · ⏱ 41 min GPT-5.6 in Claude Code and the Codex system prompt crashout – TheoTheo · 2026-07-16 Thu · ⏱ 30 min Bun's Rust Rewrite and Zig's Self-Inflicted CrashoutTheo · 2026-07-15 Wed · ⏱ 1 hr Ultra Is Not a Reasoning LevelTheo · 2026-07-14 Tue · ⏱ 25 min Using GPT-5.6 Without Burning Rate LimitsTheo · 2026-07-13 Mon · ⏱ 30 min GPT-5.6 Reviewed: Sol, Terra, LunaTheo · 2026-07-12 Sun · ⏱ 36 min OpenAI Merges Codex into ChatGPTTheo · 2026-07-11 Sat · ⏱ 26 min Burning $200K of GPT-5.6 inference across 67 agentic projectsTheo · 2026-07-10 Fri · ⏱ 26 min Grok 4.5 – a fast, cheap, token-efficient code model from XAI/CursorTheo · 2026-07-09 Thu · ⏱ 24 min Open weight is great, local models are overrated – TheoTheo · 2026-07-07 Tue · ⏱ 28 min A proper Fable 5 workflow – Codex routing, sub-agents and cost controlTheo · 2026-07-06 Mon · ⏱ 43 min Debunking the Fable 5 misconceptions – cost, nerfs and safety reroutesTheo · 2026-07-04 Sat · ⏱ 23 min Why I'm Switching to Linux (for Real)Theo · 2026-07-03 Fri · ⏱ 38 min Sonnet 5 – Agentic but Expensive, and Fable's ReturnTheo · 2026-07-01 Wed · ⏱ 28 min Why OpenAI Models Are So Token-EfficientTheo · 2026-06-30 Tue · ⏱ 31 min Apple Mac Price Hikes from the RAM CrisisTheo · 2026-06-29 Mon · ⏱ 22 min GPT-5.6 launches government-gated, and we can't use itTheo · 2026-06-27 Sat · ⏱ 30 min Why Google is falling behind in AI agentsTheo · 2026-06-26 Fri · ⏱ 19 min Claude Tag: channel-level org agents as the next paradigmTheo · 2026-06-25 Thu · ⏱ 20 min The US government ban on Anthropic's Fable 5 modelTheo · 2026-06-24 Wed · ⏱ 19 min Six things I wish someone would build: NPM, Git, mobile OS, Slack, benchmarksTheo · 2026-06-22 Mon · ⏱ 44 min Stop building glue: AI makes the boil-the-ocean horizontal play viableTheo · 2026-06-20 Sat · ⏱ 23 min Loops that prompt agents: orchestrating dynamic agent workflowsTheo · 2026-06-18 Thu · ⏱ 24 min The good parts of Claude Code, made to be stolenTheo · 2026-06-17 Wed · ⏱ 27 min Fable 5's invisible sabotage and the trust it brokeTheo · 2026-06-15 Mon · ⏱ 29 min US government bans Fable 5 over a cyber jailbreakTheo · 2026-06-13 Sat · ⏱ 14 min Tokenmaxxing Fable: Squeezing Subsidized Claude SubscriptionsTheo · 2026-06-12 Fri · ⏱ 30 min Living with Fable 5: best coding model yet, but expensive and safety-throttledTheo · 2026-06-11 Thu · ⏱ 32 min The compute crisis, visualizedTheo · 2026-06-09 Tue · ⏱ 23 min Anthropic floats pausing AI as recursive self-improvement nearsTheo · 2026-06-08 Mon · ⏱ 44 min Cloudflare buys Void Zero (Vite) to build an agent-native cloudTheo · 2026-06-06 Sat · ⏱ 30 min Laziness is a virtue LLMs can't haveTheo · 2026-06-04 Thu · ⏱ 18 min Prompts rot silently: treat AGENTS.md as model-specific tech debtTheo · 2026-06-03 Wed · ⏱ 20 min AI raises the floor for weak engineers — but widens the gapTheo · 2026-06-01 Mon · ⏱ 38 min OpenAI crushed Anthropic on a benchmark that actually matches realityTheo · 2026-05-31 Sun · ⏱ 32 min Opus 4.8 review: smarter, token-hungry, still very ClaudeTheo · 2026-05-29 Fri · ⏱ 28 min Anthropic's accidental profitability: AWS reach, tier inflation, fatter tokenizersTheo · 2026-05-28 Thu · ⏱ 27 min New AI coding workflow: GPT-5.5, Codex/T3 Code, single-thread on main, lean promptsTheo · 2026-05-27 Wed · ⏱ 47 min Claude Code vs Codex vs Cursor: a philosophy comparison, not a benchmarkTheo · 2026-05-26 Tue · ⏱ 37 min Cursor Composer 2.5: distilled Kimi K2.5 hits frontier code scores at 5% cost, but API-lockedTheo · 2026-05-24 Sun · ⏱ 37 min GitHub internal repos pwned via poisoned NX Console VS Code extensionTheo · 2026-05-21 Thu · ⏱ 27 min Google IO crash-out: Gemini 3.5 Flash, Antigravity CLI, Railway takedownTheo · 2026-05-20 Wed · ⏱ 23 min Copilot's $40 plan loophole — burning $40K of inferenceTheo · 2026-05-19 Tue · ⏱ 39 min Software security ArmageddonTheo · 2026-05-15 Fri · ⏱ 33 min Anthropic chokes Agent SDK usageTheo · 2026-05-14 Thu · ⏱ 45 min On "The Unreasonable Effectiveness of HTML"Theo · 2026-05-13 Wed · ⏱ 36 min On Bun's rewrite from Zig to RustTheo · 2026-05-12 Tue · ⏱ 27 min On "Agentic coding is a trap" postTheo · 2026-05-11 Mon · ⏱ 56 min Anthropic is leasing SpaceX's Colossus 1 superclusterTheo · 2026-05-07 Thu · ⏱ 35 min What's next? [Git forge]Theo · 2026-05-06 Wed · ⏱ 50 min Prime is right about AITheo · 2026-05-05 Tue · ⏱ 41 min MS/OpenAI break upTheo · 2026-05-04 Mon · ⏱ 36 min Anthropic's hostile-to-developer patternTheo · 2026-05-01 Fri · ⏱ 34 min I give up on GitHubTheo · 2026-04-30 Thu · ⏱ 35 min Claude Code's favorite tech stackTheo · 2026-04-29 Wed · ⏱ 38 min Career advice in the agent eraTheo · 2026-04-28 Tue · ⏱ 54 min Markdown is a terrible languageTheo · 2026-04-27 Mon · ⏱ 22 min Saving the web from Javascript bloatTheo · 2026-03-29 Sun Cursor, Claude Code, and Codex All Have a Big ProblemTheo · 2026-03-01 Sun · ⏱ 44 min Why AI Coding Tools Are Falling Into SlopTheo · 2026-02-23 Mon · ⏱ 29 min