Skills library
127 skills from 54 videos and 10 T3 Code releases. Five are new this week.
Give Sol a screenshot of something you want modeled in 3D. Tell it that it has Blender, the free 3D modeling app, available from the command line, and ask it to build the model.
In 2 videos · First seen Sep 4Run your main agent on a long goal, and keep a second chat that reads its history. You use that second chat to ask for status and what is slowing the work, which surfaces the real blockers.
In 1 video · First seen Oct 28Pi , Muse Code and OpenCode 2 support
In 0 videos · First seen Oct 15Have a coding agent rebuild an app on another platform, using the original app as the model to copy. Give it the source code, a simulator (a virtual phone on the computer) and the working app. When something is off, describe how the original behaves and ask it to match.
In 2 videos · First seen Sep 6Instead of browsing GitHub, ask an agent to list recent pull requests (proposed code changes) in the areas you care about, each with a short summary. It replaces slow, tab-heavy GitHub browsing.
In 1 video · First seen Sep 15Paste a large pull request into Claude and ask what is changing, why, and where the risk is. You can go further and have Claude build a visual artifact that teaches how that part of the code works.
In 1 video · First seen Nov 6AGENTS.md (or CLAUDE.md) is the instruction file every agent reads at the start. Theo's explains what the product is and what makes it special ('open at the core', 'performance without compromise', 'remote ready'), plus a glossary.
In 1 video · First seen Aug 25A way to write a post about a price or limit change. Lead with the bad news, state the real net effect in plain numbers, and acknowledge how users will feel it instead of spinning it as good.
In 1 video · First seen Aug 30A skill is a short markdown instruction file a model loads. Anthropic provides an official one for interface design. It steers models away from generic AI-looking pages toward designs with real ideas and nice animations.
In 3 videos · First seen Oct 8A line from Anthropic's guide: if the agent finds bugs or performance issues the task didn't mention, it should not fix them in this change, but list them as follow-ups in its summary. Anthropic says this cuts unrequested additions sharply with no drop in task success.
In 1 video · First seen Sep 22Before reviewing a large agent-made change, ask the agent how risky it is and what the worst outcome of merging it now would be. The answer tells you whether a line-by-line read is worth it.
In 1 video · First seen Sep 24Fable 5.1 stays quiet during long runs, sometimes making 30+ tool calls with no text. You get progress updates back just by asking for them in your instructions.
In 1 video · First seen Sep 3Ask one agent to describe in detail what it can do, then paste that description into another agent and ask it to clone those abilities. Theo did this to copy Codex's computer use into T3 Code using Opus.
In 1 video · First seen Oct 18Astra in OpenAI's Codex app can operate your Mac directly: clicking, opening apps and doing tasks in the background. It suits prep work, like setting up Final Cut before video editing.
In 1 video · First seen Sep 10Astra can launch many helper agents (sub-agents) that message each other and report back, and it uses their updates to keep the group on track. Fable, by contrast, plans its sub-agents in fixed steps up front.
In 1 video · First seen Sep 10Claude Code can store 'memories' about your projects. This skill is checking what it has stored, then turning the feature off, archiving the memories and deleting them. Useful when you want clean, repeatable runs.
In 1 video · First seen Aug 25First ask the model to audit the codebase and propose improvements or an ordered set of PRs (proposed code changes). Then tell it to spin up sub-agents (parallel helper agents) to do the work and merge once checks pass.
In 2 videos · First seen Sep 3A skill (saved instruction file) that tells the agent to watch its pull request: fix failing checks, update the change onto the latest code, verify and address review-bot comments, and repeat until everything is green.
In 3 videos · First seen Sep 4BAML is a small language for declaring the exact shape of data you want back from an LLM. Its runtime repairs badly formatted JSON so you always get that shape. The catch is that the repairs take time.
In 1 video · First seen Sep 20Put screenshots of rival products into the review panel, so reviewers score your app side by side with competitors instead of on its own. The goal becomes beating those apps, not just your own earlier version.
In 1 video · First seen Nov 3A setting in Claude Code that caps Haiku 5.5's auto-compact window at 100k. The session then summarizes itself before it passes the price threshold. It applies only to Haiku, sub-agents included, so Haiku sub-agents never get expensive.
In 1 video · First seen Oct 8Have a model label a large dataset and give each label a confidence score from 0 to 1. A slider sets how sure the model must be before a label counts. Theo used Jev on 32,311 messages from 1,118 of his own agent chats.
In 1 video · First seen Sep 20Anthropic gives Claude Slack channels, each focused on the health of one system component, with access to its logs and metrics. Anthropic says these catch subtle problems that normal alerts miss.
In 1 video · First seen Oct 15Haiku is Anthropic's smallest, fastest model. Theo uses it to sweep large amounts of data, or every file in a codebase, for a pattern that plain code search can't detect.
In 1 video · First seen Nov 12Opus 5.5 is Anthropic's model. Its main gain is reliability: it is less likely to do something stupid. That makes it a good fit for longer, multi-step agent workflows.
In 1 video · First seen Sep 27A small server, or middle layer, that you sign into with your Claude and Codex accounts. It keeps those logins alive and gives Claude Code and Codex one address to send requests to. It also handles juggling multiple accounts.
In 2 videos · First seen Oct 1Use your Codex subscription in CI, the automated system that checks code changes, so it reviews each pull request. OpenAI allows Codex use almost anywhere, as long as it isn't serving user traffic. Charging others for access is not allowed.
In 1 video · First seen Oct 7OpenAI's Codex coding subscription. Its $100 plans currently have no 5-hour limit, only a weekly limit, which makes it an alternative to Claude Code plans.
In 1 video · First seen Aug 30T3 Code added context compaction across harnesses (v0.0.39), which shrinks long conversation history so old threads use less. An earlier release (v0.0.34) redesigned usage insights.
In 0 videos · First seen Aug 25Connect an AI model to your email (Gmail) and notes tool (Notion) and ask it what you should be prioritizing. It can surface things you have missed or dropped.
In 1 video · First seen Sep 4ChatGPT lets you link several email accounts at once. You can then ask it about your inbox instead of reading everything yourself.
In 1 video · First seen Oct 18A skill (saved agent instructions) that sends a change, plan or design to reviewer models from several different AI companies. Each scores the work out of 10. The work isn't done until every reviewer gives a high score; Theo uses 9.5.
In 4 videos · First seen Nov 2One model writes the code and a model from a different AI lab reviews it, often in a loop until both agree. Theo has Opus write and OpenAI's GPT-6.1 Sol review. A different model family catches mistakes the first one misses.
In 7 videos · First seen Sep 22Make merging safe on two fronts. Code should be very likely fine before a person approves it. Mistakes that slip through should do little damage, because they reach a few users first and are easy to revert (undo).
In 1 video · First seen Oct 15When an agent hands research to sub-agents (helper agents it launches), tell them to read the actual code behind every claim and never cite a file they didn't open. They should also not edit anything. Muse does this on its own.
In 1 video · First seen Oct 13Require agents to attach a video or screenshot showing their change working, then watch it yourself. A new GitHub CLI update lets agents upload these without adding them to the codebase.
In 1 video · First seen Aug 24T3 Code release v0.1.8 lets you opt into environment dictation on a device, so you can speak instead of type. The notes give no more detail.
In 0 videos · First seen Nov 9Remove the 'anti-slop' writing-style rules from your agent instruction file (the CLAUDE.md the agent reads every session). Then see how the model writes on its own.
In 1 video · First seen Sep 23First build an eval (an automated test set) from real cases where agents misused your product. Get to a perfect score, then shrink prompts or tool descriptions while keeping that score. Theo cut T3 Code's tool descriptions to a third of their size this way.
In 1 video · First seen Nov 3Expo is a toolkit for React Native, a way to build mobile apps. One developer sets up a development version, then others scan a QR code and start changing it. It also makes pushing updates to users easier.
In 1 video · First seen Sep 14Using the Fable 5.2 model to review batches of pull requests (PRs), the code changes waiting to be merged. When Fable says a PR is good to go, it usually is. It skips minor nitpicks, though it misses some things.
In 1 video · First seen Nov 10Give an agent a tool that looks real, like 'run a terminal command', but never runs anything. A small, cheap model invents believable output instead. This lets you watch what the agent tries, good or bad, with nothing at risk.
In 1 video · First seen Sep 1Start loading a page's data when the user hovers over a link, before they click. Anthropic did this for conversations. It makes the app feel faster for a small change.
In 1 video · First seen Oct 5When a model keeps doing something you dislike, ask it what the pattern is called and why it does it. Then add a rule against it to your agent instructions file, AGENTS.md.
In 1 video · First seen Nov 10Run separate agent threads that watch the working agents, spot blockers and repeated failures, and change their setup. The aim is to let many agents run in parallel without colliding.
In 1 video · First seen Oct 10A repo (project folder) that documents every machine you use: how to connect over SSH (remote login), its purpose, installed tools and settings. You start agent threads from it for machine chores, including setting up new machines.
In 2 videos · First seen Sep 15T3 Code release v0.1.0 lets you fork any thread, branching a conversation into a new copy so you can try a different direction. The notes give no more detail.
In 0 videos · First seen Oct 15Instead of only building small new apps, ask the AI to copy (fork) an existing tool you use and change it to fit how you work.
In 1 video · First seen Oct 20A skill telling the agent nobody is watching. It must finish the change, prove it works, open the pull request, handle reviews and failing checks, and merge on its own. It only interrupts you if it gets stuck.
In 1 video · First seen Oct 26A version of the Full Send skill that does everything except merge. The agent keeps fixing, verifying and handling reviews until the change is ready, then stops so you make the final call.
In 1 video · First seen Oct 26A low-cost model from GLM: 15 cents per million input tokens and 50 cents per million output, often half off. It is good at long agentic tasks, and Theo says it matches or beats GPT-5 High on benchmarks.
In 2 videos · First seen Oct 11Two routes OpenAI says stay open for GPT models in Cursor after the cutoff: your own OpenAI API key (a password-like code billed to your OpenAI account), or OpenAI's Codex extension inside Cursor.
In 1 video · First seen Aug 30OpenAI's GPT-6.1 model, called Sol or Soul in the notes, used for analysis and code review rather than writing code. It tends to make few false claims at low cost, though it can miss serious issues.
In 2 videos · First seen Oct 14Give the Muse model a bug report and ask it to explain the cause and check whether a fix is already open as a PR (a proposed code change waiting to be merged).
In 1 video · First seen Oct 13After a long agent run ends with a report you can't follow, open the same thread in a different model and ask it to explain what was done. Theo uses Claude Opus 5.5 to summarize what Astra did, so he can understand the changes before trusting them.
In 1 video · First seen Nov 18Hand one model's work to a stronger model and ask for an honest critique. This catches wrong diagnoses, ignored existing fixes and missed root causes.
In 2 videos · First seen Oct 14When a post shares cost figures that seem off, screenshot it and ask Claude or Codex to check the math and assumptions.
In 1 video · First seen Oct 26When failures show up that shouldn't happen, have Claude add alerting, monitoring and tracing to find where, when and in which regions they occur, then write a report. Theo did this and uncovered serious flaws in Cloudflare's pipeline that Cloudflare didn't know about.
In 1 video · First seen Oct 30Keep your running app and the agent side by side and send many small requests, watching each change land. A fast model makes this workable, so you refine interactive parts like buttons and flows over many rounds instead of hoping for a perfect first try.
In 2 videos · First seen Oct 6Use Jev as a quick first pass for high-volume, simple sorting jobs, like clearing low-effort spam from email, before an expensive model touches them. In a clip, it averaged about 200 ms per email, 38 emails per second. It suits jobs too cheap or too frequent for an LLM.
In 1 video · First seen Sep 20just-bash is a library by Malte at Vercel that gives agents a fake bash command line in memory, with a fake file system and processes. Agents can explore files as if on a real computer without a real machine. Alone it can't build a real app, since you still need a TypeScript compiler; TS-Rust fills that gap.
In 1 video · First seen Oct 10T3 Code release v0.1.7 adds Kiro in beta, usable as an Early Access provider. The notes give no further detail.
In 0 videos · First seen Nov 4Legend List is a list library by developer Jay that draws only the rows on screen. It is written entirely in JavaScript, so it also runs on the web. It's why T3 Code's scrolling is smooth in its desktop and web apps.
In 1 video · First seen Sep 14In T3 Code, each agent thread can launch its own copy of the app, record a video of the change working, and use fake model responses so full end-to-end tests don't spend AI tokens. By the time the agent says code is ready, it has checked it most of the ways Theo would.
In 1 video · First seen Nov 6Route all AI usage through your own proxy (a middleman that relays requests to providers) and store every request and response in a database. Theo uses ClickHouse, a database for huge data volumes, on his home storage server, holding a terabyte of history. It lets him audit real costs, what works and where failures happen.
In 2 videos · First seen Oct 26Run agents autonomously overnight, not mainly to get code written but to hit failure cases. A snag agents hit once every two hours shows up two or three times in a five-hour run. Fixing those spots makes shorter, supervised runs go smoothly.
In 1 video · First seen Sep 17Give an agent a standing goal with a measurable target and tell it to keep going until it hits it, instead of a vague 'improve this'. Theo's agent kept making one speed fix and declaring victory until he set 'minimum 2x faster than the Go build', which kept it working for two weeks until it got there.
In 1 video · First seen Oct 10Luna is OpenAI's small 5.6 model, which is fast and cheap. Theo uses it for background chores in T3 Code like titles, status updates, summaries and sorting threads. It can call tools and turn messy context into a JSON object (structured data your tools can read).
In 1 video · First seen Aug 22Standing instructions for long agent runs: the agent checks its own work with computer use, runs the AI code review bots, and has sub-agents do another pass. It only reports back once it is fairly confident nothing users would notice has broken.
In 1 video · First seen Sep 15Set up a project so anyone, human or agent, can quickly start the app and try a change. Theo changed thousands of lines in T3 Code so its dev server (the local test version) runs over Tailscale, a private network, with a read-only snapshot of real data.
In 1 video · First seen Aug 24Choose models by what each job needs instead of using one model for everything. Weigh strengths like design taste, review depth or speed, and judge cost by the whole task, not the price per token.
In 4 videos · First seen Sep 23MCP (Model Context Protocol) is a standard connection that lets an agent reach data and services outside its harness, like a cookbook left in the kitchen it may choose to read. It fits when the agent needs data stored elsewhere, like your product database on Supabase or analytics on PostHog.
In 1 video · First seen Oct 30Have several models each review the same codebase and suggest areas to improve, then use strong models as judges to score the suggestions. In Theo's test, Grok 4.7 came just behind Astra on viability, and Fable 5.1 made fewer suggestions that still scored well.
In 1 video · First seen Sep 22Have the agent run every task on two models, then have a panel of judge models score results without knowing which model did which. The models can also review each other's plans. It showed Theo that Opus did more correct, deeper work while Muse did more work, faster.
In 1 video · First seen Oct 13npx t3 triage is a terminal command that hands Claude Code or Codex all the context needed to debug T3 Code on your machine, and can file an issue to the team. T3 Code connects to many tools, so odd bugs are inevitable.
In 1 video · First seen Oct 29Open-source apps that put Claude Code, Codex and other agents in one window while you pay each AI lab directly. Theo names his own T3 Code, plus Gene (from the Coolify developer), Superset and Herder.
In 1 video · First seen Aug 30From one thread, ask the agent to start helper agents on different models that work on the same task. T3 Code's delegate_task lets an agent start child agents on any provider or model and wait for their results. Each helper runs in its official tool on a subscription you already have.
In 1 video · First seen Oct 15An automatic check on every pull request (a proposed code change) that measures something you care about, like how much data is sent. It posts the numbers on the PR and flags or fails regressions. Theo's replays real thread data, after small changes kept undoing his speed work.
In 3 videos · First seen Aug 25Effort level controls how long the model thinks and checks before answering. Opus 5.5 offers low, medium, high, extra-high and max. You can change the level mid-session without breaking the cache, the reused earlier context that keeps long sessions cheaper.
In 2 videos · First seen Sep 23pnpm, a package installer, can use hard links: many file names point to one stored copy instead of duplicating files. The result depends on the file system underneath. On Linux XFS one reinstall test fell from 7 to 2.85 seconds; on a Mac installs grew from about 40 to about 100 seconds.
In 1 video · First seen Aug 20A skill (saved instructions for an agent) that hands someone else's pull request to a fresh agent. The agent reviews it, decides whether to build on it or use it only as a reference, and manages the branch until it lands. The original author gets credit.
In 2 videos · First seen Sep 3Point a cheap model at all open pull requests and have it sort them, for example into merge, close or needs another pass, or by ease of merge and value to users. One run cost Theo about 12 cents for a useful insight.
In 3 videos · First seen Aug 22For 3D or interactive projects, Astra builds the first version to get the look roughly right. Then Fable does the detail work on animation curves, camera movement and overall feel.
In 1 video · First seen Sep 10Delete automated tests that add no value. When a pointless test shows up again, ask the agent that wrote it why, and fix that root cause. Useless tests multiply and breed more pointless edge-case tests.
In 1 video · First seen Nov 9T3 Code has a pull requests page that works across providers, with in-app reviews. Releases added sending PR line requests to the agent, easier-to-scan comments, guided reviews (a beta tour of a PR) and @user mentions in PR comments.
In 0 videos · First seen Aug 25Keep a set of prompts that earlier models failed at. When a new frontier model comes out, run that set before your usual prompts. It shows what the new model can now do.
In 1 video · First seen Sep 17Have agents build a narrow, custom version of a paid tool that fits only your needs. Theo had agents copy his analytics data from a provider into his own ClickHouse database (a fast analytics store) on R2 (Cloudflare file storage).
In 1 video · First seen Oct 15When Astra loops on a pull request (a proposed code change) without finishing, stop it and give the PR to Fable 5.1 to finish and land cleanly.
In 1 video · First seen Sep 10T3 Code lets you rewind a conversation to an earlier point while keeping the file changes already made. Added in release v0.0.42.
In 0 videos · First seen Sep 15Rift by Anomaly is an alternative to git work trees for Mac users. It virtualizes files so the agent sees a full copy, and only writes new files when they change (copy-on-write). Theo says it makes work trees instant on macOS.
In 1 video · First seen Aug 20T3 Code added scheduled tasks in release v0.1.0. Since v0.1.8, each task links to its thread.
In 0 videos · First seen Oct 15Instead of looping until reviewers have no comments, each review agent scores the change, and you set the score a PR must reach. Scores can track things you care about, like how likely you are to be okay with it, or several scores at once.
In 2 videos · First seen Nov 6A way to make a fuzzy goal like "looks as good as the real game" checkable. After each good build, the agent screenshots set moments, compares them with reference footage, and a panel of AI judges says whether it looks good enough.
In 1 video · First seen Oct 28Paste a screenshot of a user's bug to the agent and have it fix the bug, test it, record a video proving it works, open a PR (a proposed code change) and handle review comments. The agent carries the job from vague screenshot to working PR.
In 1 video · First seen Sep 17Use the agent instructions file (AGENTS.md, standing instructions the agent reads every session) to set how it writes. Theo's says he has ADHD, asks for short, easy-to-read answers, varied sentence length, and simplified technical English.
In 1 video · First seen Oct 13Configure machines around what your agents need, not your own preferences. Theo is building T3OS, a customized Ubuntu image for his agents. He found the Zsh shell (command-line environment) caused over 30% of Claude's errors, because its newline handling garbled search commands.
In 1 video · First seen Oct 20In T3 Code you can share a thread with a guest. The guest watches live and can send messages, which you approve. Added in release v0.1.3.
In 0 videos · First seen Oct 27Side chats in T3 Code sit beside the main thread. Since release v0.1.8 you can start one by typing /side.
In 0 videos · First seen Oct 31To run an AI model on your own machine, the whole model must fit in the graphics chip's memory. Splitting a model across two consumer graphics cards doesn't really work, because the link between them is far too slow to share memory.
In 1 video · First seen Aug 28Keep a smart model like Opus as the main agent and let it hand small jobs to cheaper, faster sub-agents (helper agents it launches itself), such as Haiku 5.5 or Sonnet 5.5. The main model plans and directs; the helpers do the many small attempts or investigations in parallel.
In 2 videos · First seen Sep 29Write a detailed bug report for the agent: what is wrong and what the fix should look like. Give it even a minimal way to check its own fix.
In 1 video · First seen Aug 24Give each AI provider's subscription a specific job instead of using one for everything. Theo uses OpenAI for computer use and verification, Claude for day-to-day coding, and Grok only for its Grok bot assistant.
In 1 video · First seen Nov 2link multiple pull requests to threads
In 0 videos · First seen Sep 15One command hands a written support playbook to Claude Code or Codex. The agent asks what went wrong, gathers reproduction steps, checks the code, then fixes the problem or files a clean bug report with secrets removed.
In 1 video · First seen Oct 18T3 Code is Theo's free, open-source app for running coding agents. It sits on top of Claude Code and Codex, using your existing subscription, skills and plugins. Its sidebar works like an email inbox for many parallel agent threads, which can run on machines you pick.
In 3 videos · First seen Oct 1T3 Connect lets you install an agent runner on any spare computer, such as an old Linux box, with one command. You then manage that machine's agent threads from the T3 Code desktop or phone app. It is free for three devices.
In 1 video · First seen Oct 29T3 Code release v0.1.2 (October 2026) added a t3 thread command. With it you can start, message, and read T3 Code threads from the terminal, the text window where commands are typed, without opening the app.
In 0 videos · First seen Oct 22Add a line to your agent's skill (a reusable instruction file) telling it to stop waiting on a slow helper model when other work remains. Theo did this when his agents used Gemini as a 'judge' and it was too slow.
In 1 video · First seen Nov 13A way to try any pull request (proposed code change) yourself, even when agents run on a remote machine. An agent serves the PR's dev server over Tailscale (a private network between your devices), and a PR label triggers a downloadable app build.
In 1 video · First seen Sep 8Ask for everything in one prompt: build the feature, file a PR (proposed code change), spin up a test copy you can try, and babysit the PR by watching and fixing issues. This avoids a back-and-forth each time you return to the thread.
In 1 video · First seen Oct 1A line added to prompts that lets Opus push back: 'If you think this is a good idea, do it. Otherwise, suggest something else.' Opus rarely challenges the task it's given, so this invites it to.
In 1 video · First seen Nov 12A way to judge how bad a security bug is: how long until everyone actually runs the fix, and how much access the attack gives. A bug in your own server is fixed once; one in installed software or a library stays open until everyone updates.
In 1 video · First seen Oct 22After a first round of fixes stalls, ask the agent to dig deep and return several extreme proposals, explicitly allowing it to "boil the ocean". It is for breaking through when normal fixes stop helping.
In 1 video · First seen Oct 5A prompt structure: state the goal as non-negotiable, offer your idea 'in my opinion' so the model can push back, admit what you don't understand, and ask for an audit rather than code, with permission to stop early on a simple fix.
In 1 video · First seen Sep 8Swing between building features and improving the foundation. When a feature is harder than it should be, stop, make the codebase easier to build on, then return to the feature.
In 1 video · First seen Nov 9A hands-on way to do performance work: name one slow thing, have the agent list theories, build demos of each, then try them yourself to see if it actually feels faster.
In 1 video · First seen Oct 5Every long task prompt states what you want, what counts as done, and when the agent should stop and ask you. Models are trained not to give up, so they tend to answer their own questions and push on.
In 1 video · First seen Sep 24A way of setting up your codebase so a very capable developer who has never seen it can do useful work right away. Each agent session starts with no memory of earlier ones, so every thread is effectively a new hire.
In 1 video · First seen Sep 6Tool calls are the actions an agent takes besides writing, such as git commands, builds or browser snapshots. When a fast model writes in 30 seconds instead of 10 minutes, these slow steps can double the total wait, so it pays to cut them.
In 1 video · First seen Oct 6Collect every time your coding agents failed, such as a broken tool call or a wrong path through your code. Turn each one into an eval, a repeatable test of whether the agent gets it right, so you can measure how well your codebase steers agents.
In 1 video · First seen Nov 5Before a big change, ask agents to review your architecture or planned changes and to investigate theories you want to test. You keep the strategic decisions and use the agents as advisors. Once you have a confident plan and a way to check the result, they carry it out quickly.
In 1 video · First seen Aug 24When you take over code nobody remembers, ask an AI questions about it. Have it check its answers against how the code actually behaves. This helps you get up to speed on old code.
In 1 video · First seen Sep 6The only note here is T3 Code release v0.1.8 (November 2026). It lets you @-mention a model to name it for the agent. The notes do not describe how to use T3 Code from other apps.
In 0 videos · First seen Nov 9Set up extra agents on crons (timed, recurring jobs) that check whether a long-running agent has stalled waiting on a question. They notice when the main agent is stuck so it can be unblocked.
In 1 video · First seen Nov 18CLAUDE.md is the instruction file Claude Code reads in every session. Anthropic suggests a rule for it: keep going when no input is needed, and stop only when blocked or before anything destructive. This cuts needless pauses on long runs.
In 1 video · First seen Sep 24Wispr Flow turns speech into text on screen. Theo pairs it with a small desk podium mic, about $70, that picks up even whispers, so he can dictate quietly around other people.
In 1 video · First seen Sep 15Format a Linux drive as XFS with VDO underneath. VDO is a layer that compresses data before it reaches the disk, and LZ4 is a lightweight compressor that keeps reads and writes cheap. Duplicate files across work trees (separate copies of a repo for parallel agent work) then take little real space.
In 1 video · First seen Aug 20Theo Daily is unofficial and not affiliated with Theo or t3.gg. Written by Claude Opus 5.5 from each video's transcript.