https://halwright.tv/2026-10-12 https://halwright.tv/start2026-10-12 https://halwright.tv/s/agents-build-a-live-show2026-10-08 https://halwright.tv/s/build-with-an-agent2026-10-12 https://halwright.tv/e/agent-fixes-a-web-server2026-09-08https://i.ytimg.com/vi/vj-ZNQdXzRo/maxresdefault.jpgAn agent fixed our web server. I never typed a command.A web page on one of the lab's servers goes down, and the problem goes to an agent in plain language. It connects over SSH, finds the failed service, traces the logs back to what stopped it, restarts it, and checks the page from outside the machine. Nobody typed a command on the server.https://www.youtube.com/embed/vj-ZNQdXzRo1252026-09-08T00:23:17Zyes https://halwright.tv/e/domain-experience2026-09-08https://i.ytimg.com/vi/LbmiXxwMsj4/maxresdefault.jpgWhy domain experience is the whole gameThe idea the channel is built on: once an agent can handle how to build or fix something, the valuable part is knowing what should exist, what broken looks like, and what fixed has to mean. A short talk piece with one worked example and the case for verification judgment as what makes an agent's speed safe.https://www.youtube.com/embed/LbmiXxwMsj42562026-09-08T00:23:17Zyes https://halwright.tv/e/whos-behind-hal-wright2026-09-08https://i.ytimg.com/vi/ykASTccNWpk/maxresdefault.jpgWho's behind Hal Wright? Meet John.Hal Wright is a pen name with a synthetic voice. This is the one video about the real person behind it: where the experience comes from, and why that experience is the premise of everything else on the channel.https://www.youtube.com/embed/ykASTccNWpk2202026-09-08T00:23:16Zyes https://halwright.tv/e/the-plan-before-the-vm2026-09-15https://i.ytimg.com/vi/SP2PwnhqeDY/maxresdefault.jpgThe plan before the VM: 27 hours, 15 artifacts, not one machine turned onBefore any machine was turned on, the build was a folder of planning documents: about twenty-seven hours of work and fifteen artifacts. This walks through each one and what it settled, from the build sheet and the decisions register to the cast, the mock episode, and the audition harness, and explains why the build starts without a rehearsal.https://www.youtube.com/embed/SP2PwnhqeDY7672026-09-15T02:42:59Zyes https://halwright.tv/e/google-cloud-project-for-an-ai-agent2026-09-25https://i.ytimg.com/vi/xu1tYknGOjA/maxresdefault.jpgGiving an AI agent a Google Cloud project without handing it the keysThe first evening of the build, on camera. An agent creates a cloud project in its own folder and on its own billing, takes an identity it borrows through service-account impersonation instead of holding a key, and sets a budget that detaches billing at the cap, proven with a synthetic alert first. It also covers the setup lessons that cost the most time: which shell the agent works in, and why credentials live in a context.https://www.youtube.com/embed/xu1tYknGOjA5112026-09-25T00:32:59Zyes https://halwright.tv/e/dnssec-on-google-cloud-dns2026-09-28https://i.ytimg.com/vi/lHH9nSuXuK0/maxresdefault.jpgAn AI agent sets up DNSSEC on Google Cloud DNS, start to finishAn agent moves two new domains onto Google Cloud DNS with DNSSEC on from the start, then stops where its reach ends: the registrar account is personal, so a second agent sets the nameservers and DS records there with a human approving each save. Every step is then proven from outside the system, from the delegation and signatures to a full DNSViz chain of trust, and the zone gets its first real record. It also covers why the registrar's one-click DNSSEC toggle is never used.https://www.youtube.com/embed/lHH9nSuXuK05072026-09-28T15:00:31Zyes https://halwright.tv/e/protecting-an-agents-inbox2026-09-27https://i.ytimg.com/vi/bWE9WLMin5w/maxresdefault.jpgProtecting an AI agent's inbox from spam floods and prompt injectionThe show takes letters from the public, and every letter a model reads costs money. This builds that front door and then attacks it: the public's input is walled off, a scanner reads every letter before any model does and is tested against a deck of hostile letters, a human check guards the form, and 1,500 letters are fired at the mailbox until storm mode trips and nothing reaches a model. The design is shown; the detection rules and thresholds are not.https://www.youtube.com/embed/bWE9WLMin5w6032026-09-27T19:10:23Zyes https://halwright.tv/e/your-job-when-agents-do-the-work2026-09-23https://i.ytimg.com/vi/9v5BnezZYSE/maxresdefault.jpgWhen AI agents do the work, what's your job?Over five working sessions an agent stood up a complete system on cloud infrastructure, and the operator never opened the virtual machine. This talk piece covers what the operator's job became instead: holding the credential boundary, steering at the load-bearing forks, challenging "done," inspecting what you expect, delegating overnight, and keeping the record that lets the work stop and resume.https://www.youtube.com/embed/9v5BnezZYSE5032026-09-23T04:12:37Zyes https://halwright.tv/e/securing-an-agent-built-system2026-09-23https://i.ytimg.com/vi/ICSy7eK-r60/maxresdefault.jpgSecuring an AI-agent system on Google Cloud, layer by layerThe security of an agent-built cloud system, in the order it was built: the box and its blast radius, one key per purpose, budgets that actually stop spending, and a network with no way in and one way out. It also covers why permission errors in testing mean the design is working, what a public surface really exposes, and why recovery is part of security.https://www.youtube.com/embed/ICSy7eK-r605402026-09-23T04:47:40Zyes https://halwright.tv/e/the-finished-system2026-09-24https://i.ytimg.com/vi/3v-QB-oiC6Q/maxresdefault.jpgAI agents built a YouTube show that publishes its own episodes. Here's the systemSix working sessions after the plan, the system argues each question live and publishes the show to its own channel. This wrap-up covers the final security sitting, the component that was never built, a kill switch tested by actually revoking the keys, what the first episode cost, and review before the upload.https://www.youtube.com/embed/3v-QB-oiC6Q3962026-09-24T21:53:38Zyes https://halwright.tv/e/the-day-after-done2026-09-30https://i.ytimg.com/vi/r8Vnx-sv7Fw/maxresdefault.jpgMy AI agents said the build was done. Here's what was missing.The build was declared finished, and the first real request sent through it found that the website's front door had never been wired. This is the day after done: a frontier model audits what the helper agents dropped, the trace turns up six gaps, one question uncovers a missing feature, a pick goes wrong and gets rolled back, and the first viewer ticket goes end to end. The next morning the schedule ran with nobody touching it.https://www.youtube.com/embed/r8Vnx-sv7Fw5102026-09-30T02:16:21Zyes https://halwright.tv/e/a-plan-is-a-budget-for-change2026-10-08https://i.ytimg.com/vi/0OKWNw1zhLo/maxresdefault.jpgA plan is a budget for change: what 27 hours of planning bought an AI-agent buildWhat twenty-seven hours of planning bought the build, and what to do on the day the plan breaks anyway. A talk piece on plans as budgets for change: plan in the big model and execute in the middleweight, size every phase to one context window, write the irreversible moments into the plan so the agent stops and the human decides, and the week a helper agent quietly kept a prototype nobody asked for. Plans change; quiet changes don't.https://www.youtube.com/embed/0OKWNw1zhLo4472026-10-08T07:56:40Zyes https://halwright.tv/e/agent-built-my-dictation-app2026-10-05https://i.ytimg.com/vi/yV0N-9Z-U0I/maxresdefault.jpgAn AI agent built my dictation app in one afternoon. The plan did the work.The commercial dictation tool kept writing Hal as howl, so an agent built a replacement in one afternoon: local Whisper speech-to-text, a classifier that settles misheard words, a small language model only when the classifier is unsure, and a dictionary that learns from edits. The point of the story is not the code. It is the plan: seventeen lines of wishes, one planning interview until every question was closed, homework only a human can do, measurements before any code, then six steps each tested by using the app.https://www.youtube.com/embed/yV0N-9Z-U0I5712026-10-05T07:55:33Zyes https://halwright.tv/e/agent-got-my-app-signed-instead2026-10-12https://i.ytimg.com/vi/H-EQ9n4flSI/maxresdefault.jpgI came in to rewrite my app. The agent got it signed instead.The dictation app from the previous piece would not run on a corporate PC: unsigned, it looks like spyware. The human came back planning a Python rewrite; the planning conversation priced the real problem, signing, and found that the Microsoft Store signs a package for free and the developer account now costs nothing. On camera: the brief dictated with the tool itself, the fee that was zero, a packaging test that lost the database and produced export/import, a measurement that kept the older correction model, the human-only steps routed through a two-agent ticket relay, and the email from Microsoft two hours later. The lesson: state the challenge and the goal, not the method.https://www.youtube.com/embed/H-EQ9n4flSI4982026-10-12T07:55:51Zyes