Peter Yang spent most of 2026 defaulting to ChatGPT because Claude Opus had become judgmental and started every answer with “here’s the honest truth.” He put the new Opus through five practical tests on his own channel, from 3D scene generation to editing his videos, and ends with the personality prompt that Opus 5 bombed. ...
Murali Swaminathan is the CTO of Freshworks — 15 years old, publicly traded, ~4,500 people — and he spent this hour (58 min) describing what it takes to rebuild that org for agents while the plane is still flying. Almost none of it is about model capability. It’s about the machinery around the model: review budgets, confidence thresholds, access control, and pricing that survives agents replacing the seats you used to bill for. ...
Ethan Bloch runs product for ChatGPT Finances at OpenAI. He founded Digit (the SMS money-saving chatbot, acquired 2021) and a second company that OpenAI acquired in April 2026, and he walks Peter Yang through six ways to actually use the product — then talks about how his team ships. ...
Andrew Warner and Adam Brakhane run through the week’s ten trending GitHub repos in 26 minutes, and the list is mostly guardrails: something that reviews the code your agent wrote, something that makes its answers short enough to read, something that keeps its context clean, and two skill packs that hand it an engineering process instead of hoping it has one. The top 10 Open Code Review (36.6k stars) — Alibaba’s code-review tool, number one this week. Deterministic pipelines plus an LLM agent, line-level comments, and built-in rules for the classic bug classes (null pointers, thread safety, XSS, SQL injection). It ships a CLI and agent skills, and Andrew’s pitch is wiring it into a Git hook so every commit gets checked before it lands. The demo he cites: someone deliberately planted a plaintext password, and the tool found it, then produced a full HTML report from one command. I Have ADHD (48k stars) — still growing, and the hosts keep coming back to it. One skill file that tells your coding agent to keep answers short and scannable. Andrew shows a before-and-after from the Claude web app where a slash command turns a wall of text into something readable — and Adam’s reason it keeps mattering is that the models themselves are getting more verbose. God’s Eye View (37.8k stars) — open data on a photorealistic 3D globe: TomTom traffic, public CCTV cameras you can click into live, aircraft overhead. Also distributed as a plain installable app, so you don’t have to assemble it from source. ECC (262k stars) — the biggest number on the list. A workflow pack that encodes planning, test-driven development and review as commands you invoke rather than advice you remember. Adam’s framing: it injects decades of engineering habits into a session, which matters most to people who never picked those habits up. Context Mode (23.5k stars) — a sandbox for tool output. Instead of your main session loading fifteen files to answer one question, a side session does the searching and hands back the answer, and that noisy conversation gets discarded. The advertised figure is 98% less context clutter; the caveat the hosts add is the useful part — compress something the model actually needed and it re-requests the full context, which is expensive on a long conversation. WeKnora (27k stars) — Tencent’s knowledge platform. Point it at your files and it builds a queryable RAG, a self-maintaining wiki and a knowledge graph. Adam’s description is the clearest one on the episode: Obsidian plus RAG plus AI plus a graph database, where the links between people, meetings and projects get built for you. OpenAI Plugins (7k stars) — 62 plugins written for Codex, each bundling an integration together with the skills and scripts that make the integration useful. Claude will either convert them or bridge them through a connector, and Adam’s advice is to install from the platform you actually use. Worktrunk (8k stars) — a CLI for Git worktrees, the feature that gives you several checked-out copies of the same project side by side on disk. Creating, listing, merging and deleting a worktree each become one short command, which is what makes running several agents on the same repo practical. Agent Skills (96.4k stars) — Addy Osmani’s pack of 25 skills that the hosts single out because it is one ordered workflow (spec, plan, build, test, review, ship) rather than a pile of unrelated files. Adam’s point: most people skip straight to build, and test and review are the two steps that get dropped. Claude Code (146.3k stars) — the repo exists, the code does not. Adam’s read: it gives people a normal place to file bugs and lets Anthropic track issues internally, so it is a repository minus the source. The skills vote Last week the hosts asked viewers whether to keep covering skills at all. The result was 32 keep to 9 kill, with strong feelings on both sides, and Warner says he spent a Saturday trying to answer the comments individually. The argument that moved the conversation: skills matter most for the less capable models. A frontier model already knows how to run test-driven development; a local or open-weight model may need the steps spelled out. Warner’s middle position: even a skill he never installs shows him the creator’s mental model. He also cut a separate video on the most-installed skills according to Vercel, whose CLI is how many of them get installed. The verdict for the format: skills stay in the roundup while they keep trending. Plugin security Adam’s distinction: a skill is text and the model stays in charge. A plugin can ship scripts and code, and installing one means handing over your files, your source and your keys. His rule: install plugins from the vendors — OpenAI or Anthropic — and treat third-party plugin marketplaces the way you treat installing unfamiliar software. Viewer submissions Logic Loop (12 stars) — one workspace for several concurrent agent sessions. The premise on screen: with five agents running, the thing that breaks is your memory, not theirs. Claude Dashboard (5 stars) — a local board for every Claude Code session on a machine: live sessions, transcripts, costs, git status, and which one is stuck. Deck Gauge (4 stars) — one board over Jira, GitHub, GitLab and Azure DevOps with DORA metrics. Warner pushed back on pulling tasks out of Jira, then made the case for it anyway. Campus AIOS (10 stars) — an AI staff in one folder: a chief of staff plus departments that remember your business. Reckoner (47 stars) — a credits and billing dashboard across AI providers, still climbing since it got its first star on the show last week. “Why not just give each of those agents its own little playground where it can have its own copy of the code, test it itself, edit it, and then once the feature is done, merge all of them back into your one main branch?” — Adam on Git worktrees ...
Peter Yang walks through ten things he actually does with Muse, Meta’s consumer personal agent, a week after launch. Two of them have a dollar figure attached, and both come from the same trick: the agent does the research and the phone calls, the human does the irreversible bit. (17 minutes, his channel.) ...
TwoSetAI Workshop #5. Ahmet İlten, founding engineer at Sentience (agent architecture and quality), walks through their personal-agent product and the four principles his team arrived at the hard way. Benjamin Carsley was scheduled to cover the voice/tone half but was out sick, so the session is Ahmet solo plus Q&A with host Angelina Yang. (48 minutes.) Sentience is a New York startup founded in 2025 by Sam Kececi, backed by Bain Capital Ventures and South Park Commons; it left closed beta the week after this recording. ...
Ali Parandeh — a chartered mechanical and software engineer, O’Reilly author (his book covers building generative AI services with FastAPI), previously head of engineering and now running his own AI advisory for heavy industry — interviewed on TwoSetAI (30 min). His market is construction, automotive, energy, aerospace and infrastructure: sectors where a bad model is not a bad decision, it is a physical failure. He came back to the show to talk about how those companies actually buy AI, how to scope projects that cannot be allowed to fail, and the job title his business model ends up needing. ...
Cole Medin (16 min) is a year into his personal AI second brain — an agent that builds a knowledge base as he works — and this one answers the question he says he keeps getting: how do you turn a second brain into a team brain? ...
The Next New Thing’s Andrew Warner and Robbie Abed (founder of NoMiddle) go through the ten most-installed agent skills on GitHub, ranked by installs through Vercel’s skills CLI. It is a short episode — twelve minutes — and four of the ten entries are Matt Pocock’s, which is its own finding: a handful of small text packages are becoming the default instruction layer for coding agents. The top 10 find-skills — the most-installed skill is a skill for finding skills. You ask your agent for something (“find a skill for writing tests”) and it comes back with candidates to install. Warner’s framing: the community needed an app store and built one as its first app. grill-me — a few lines that invoke the grilling skill. Warner ran it to prep for the episode and it opened by asking what he was trying to convince the listeners of, which he calls the endgame question, asked before any work started. frontend-design — Anthropic’s skill for stopping the signature AI look, starting with that warm cream background. The episode shows the same site built before and after: same request, different color logic and more deliberate layout. agent-browser — headless browser control for an agent, no visible mouse. The examples on screen are the practical ones: log into the dashboard and screenshot the billing page; open staging, sign up as a test user, and tell me if it works. setup-matt-pocock-skills — a one-time scaffold that wires his engineering skills into a repo and configures the things those skills assume: issue tracker, triage labels for the five canonical roles, where domain docs and ADRs live, and how consumers should read them. Robbie’s read: it is the first 60 days of a consultant, delivered as setup. vercel-react-best-practices — roughly 70 React and Next.js performance rules handed to the agent as rules. You do not have to understand all of them for the skill to be worth installing, since the point is the mistakes the agent now avoids. grilling — the engine behind grill-me, split out so it can be reused anywhere the agent needs to interrogate you: how to sequence the questions, how to read the answers, what to push on. Both hosts treat it as the same idea in a reusable container. lark-doc — ByteDance’s workplace platform, Feishu outside China and Lark inside: contacts, messenger, the spreadsheet equivalent, calendar, wiki, video conferencing. Warner is candid that almost nobody watching will use it — it is on the list because the user base is enormous. teach — turns the agent into a tutor that accumulates lessons across sessions as numbered HTML files in a lessons folder. Matt’s demo teaches him to solve a Rubik’s cube; the interesting part is that it remembers where he stopped and what he stumbled on, and builds the next lesson on both. web-design-guidelines — Vercel’s web standards as an agent checklist: icon-only buttons need an aria-label, form controls need labels and autocomplete, plus typography and animation rules. The demo has the agent review a checkout page and report that the icon button has no accessible name. Where the list is really coming from Vercel shows up four times (find-skills, agent-browser, React best practices, web design guidelines) and Matt Pocock four times (grill-me, setup, grilling, teach). Two authors account for eight of the ten slots. The grill-me / grilling split is the pattern to notice: one skill is the entry point (“grill me before I build”), the other is the reusable engine underneath it. Install counts measure the install path as much as the content — Vercel’s CLI is the thing people install skills through, so its own skills are the easiest to install. Skills are readable. The hosts keep pointing out that you can open the file and read plain English instructions, which is what makes them inspectable rather than magic. Where two skills disagree frontend-design and web-design-guidelines can pull in opposite directions: one optimizes for a distinctive, animated interface, the other for accessibility compliance. Robbie’s example is a healthcare clinic — the animation that reads as tasteful on a portfolio site fails an accessibility review, and the two installed skills will argue about it through your agent. The practical upshot: skills stack, and stacking them means owning the conflicts rather than assuming they compose. The takeaway Ranked by installs, the top of the list is not about making the model smarter. It is infrastructure around the model: discovering other skills, interrogating a plan before it is built, keeping the codebase’s rules in front of the agent, and controlling a browser or a workspace. The two that lean hardest on taste — frontend-design and web-design-guidelines — are also the two most likely to contradict each other, which is a fair description of where agent tooling is right now. ...
Peter Yang walks through the ChatGPT Skills chain that runs his podcast production — 15 manual steps a week, three skills, one orchestrator. (19 minutes, his own channel; prompt pack at behindthecraft.com.) The three-step system Map every manual step. He wrote out all 15 things he used to do per episode — research the guest, build the interview guide, send edit instructions, transcript, upload, title/thumbnail, newsletter, clips, social posts. Just making the list exposed three distinct phases hiding inside it Build one skill per phase — podcast prep, podcast edit, podcast production. Production is the orchestrator: it walks him through thumbnail, title, show notes, newsletter, social post and clips one by one, calling the individual skills Connect the skills end to end. His framing: if you find yourself repeating the same template or process, build a skill for it The first move after the list is to paste the list into ChatGPT and ask “what work from here can you take off my plate?” He runs that in a custom mode so his existing skills don’t shape the answer — he wants to see what the model claims before it sees his setup. ...
Hamel Husain and Isaac Flath run a model’s writing through the cheapest eval there is: read the drafts out loud and argue about them. (Two-minute clip from the same “Writing Evals With Isaac” live session.) The task Isaac’s blog post describes a document QA pipeline built over a 77-page set with 22 questions. Every answer cites a bounding box drawn on the exact spot on the page, so checking an answer takes seconds instead of a full re-read They hand the post to a model and ask for an X post about it, focused on the pieces of the pipeline that can be tested Three drafts, one task, different prompts — same three candidates as the session’s writing-style comparison What they liked A real number, early and concrete: the model answers 197 for the underwriting fee — and “if you have to reread the whole document to find out, the answer saved you nothing” The pipeline stated as testable stages with the reason for splitting it: once checking is cheap, experiments are cheap, so you can swap one stage at a time and rerun the same 22 questions One fewer fact. The winning drafts carried less than the others, and that was an improvement rather than a loss Precision in naming: “finding (grep vs semantic search), cheapest run” beats “GPT vs semantic search over the same OCR output, same accuracy, cheapest run” — the longer version buries the point that the bottleneck was never search What they didn’t “Very similar but a little bit more wordy” is the entire verdict on one draft Drafts that pull the pipeline’s three steps out of the source post without explaining what they are — accurate and unreadable An extra fact about cost that added nothing to the argument Editorial framing that tells the reader what the result means before showing it The rule the description writes down An AI draft can get the facts right and still not be good. The ones we liked had a real number in them and one less fact — turns out cutting helped more than adding. ...
Hamel Husain and Isaac Flath take Anthropic’s new anti-slop writing guidance and put it through an actual comparison: one question, three system prompts, and a verdict reached by reading the drafts out loud. (Seven-minute clip from their “Writing Evals With Isaac” live session.) The thing being tested Anthropic published a named anti-pattern for its newest writing model, Claude Fable 5.1: mannered prose — metaphor and flourish substituted for direct statement The prompting guide treats it as a model-specific behavior. Fable 5.1’s writing is a step up from earlier Claude models (fewer stock phrases, less unexplained jargon), but in some cases it runs denser than Fable 5 — longer sentences, fewer paragraph breaks, more figurative phrasing standing in for plain statement The fix is one paragraph, pasted into a system prompt: Mannered prose substitutes metaphor and flourish for direct statement. Instead of “a parameter worth varying,” the mannered writer produces “a dial worth turning.” Instead of “this point still matters,” they write “this point earns its keep.” The phrases exist to display the writer, not to convey the idea, and readers can tell. That is why mannered prose irritates: it makes the reader work harder so the writer can perform. It is also imprecise. Metaphors drag in connotations the writer did not choose and cannot control. The fix is to say what you mean. When a literal phrase is available, use it. ...
Cole Medin walks through “drive screen” — a coding-agent skill that lets Claude Code or Codex drive his entire desktop using nothing but shell commands. 13 minutes, no harness, no dependencies. The pitch: computer use without the harness Cole says he was a computer-use skeptic: tools like Claude’s built-in computer use, Codex computer use, and the open-source options were “bloated and hard to manage” — hard to customize as little failure modes show up. His claim: with frontier models (Fable 5.1, GPT-6 Astra), an agent can drive the screen reliably through PowerShell on Windows, AppleScript on Mac, and plain shell on Linux. Nothing gets installed. Tested across Mac, Linux, and Windows, plus multiple monitors and display settings. The skill is a little under 400 lines. The broader takeaway he keeps returning to: tasks he had written off as “the LLM can’t do that” (computer use, editing video) no longer need a specialized tool — just try it. What he actually uses it for Morning setup — one prompt (“get my morning set up, here’s what I’m working on”) opens browser tabs, Obsidian notes, desktop apps, and Docker containers. He estimates 10–15 minutes saved daily. Staging demos and recordings every day. Testing desktop apps — the skill was born from wanting to try open-source projects (e.g. Kestrel) that have a UI and can’t be exercised through browser automation. The first test: hand Claude a GitHub URL, tell it to research the repo, launch the app, and drive the screen to test features, then leave it open. It worked end to end. Driving other agent sessions — watching for approval prompts and auto-driving them. How the skill is built It starts by asking whether screen control is needed at all. The skill explicitly tells the agent to push back, because screen control is the slowest and least reliable way to make a computer do something — browser automation is often the better answer. Ships with a custom CLI of scripts for window discovery, focusing, typing, and pasting — packaging his hours of testing into deterministic commands instead of having the agent improvise shell syntax in real time. This is the main reliability win. Hard rules — lessons learned the hard way, encoded. A control loop — discover the window, screenshot it, focus, then act. Traps — failure patterns he kept hitting with browser tabs, desktop apps, and other agent sessions. You can delete or add your own. Security and prompt injection The obvious risk: indirect prompt injection arriving on screen, which the agent reads and acts on. He links HiddenLayer’s writeup on indirect prompt injection against Claude computer use. His position, which he admits may age badly: with the newest models, prompt injection is much less of a concern than it used to be — and he avoids using the skill for anything production-grade, keeping it to simple tasks like morning setup and desktop-app testing. Anything production still gets a full harness. He also notes the skill is token-efficient: the slowness is waiting on screen input, not burning tens of thousands of tokens. Install Skill folder: github.com/coleam00/skills under .claude/skills/drive-screen — drop it into any project, hand the URL to your coding agent, or take the ideas and install nothing. “Screen control is the slowest and least reliable way to make a computer do something. So it’s the most adaptable and flexible, but it’s the slowest.” ...
Olga Beregovaya — VP of AI at Smartling, who entered natural language processing in 1997 when rule-based machine translation still ruled — interviewed by Angelina on TwoSetAI (73 min). She has spent 25+ years watching the discipline get rebuilt, and thinks the next thing to go is the source text itself. ...
Pedro Franceschi (co-founder and CEO of Brex) on Peter Yang’s channel — 48 minutes, two live demos: Brex’s AI recruiting employee, and the OpenClaw harness Pedro runs his own life on. All data shown is Brex demo data. Build employees, not agents The framing: don’t build an open-ended agent that could do a million things — build a virtual employee with a specific job. The test is whether it behaves, looks and feels like a real person. Jim, Brex’s AI recruiter: sources candidates, filters all inbound, and serves as the recruiting analytics layer. Running since February. Three-stage pipeline: sync Greenhouse (the ATS) → ingest candidates (resume, LinkedIn profile, GitHub data) → evaluate every candidate against role-specific criteria. Recruiters talk to Jim in Slack — “which role has the highest average candidate score?”, “top 30 applicants for this role” — and it posts standout candidates with links. Pedro’s point: it behaves like a coworker. A self-improvement loop: Jim proposes things to build that would automate more of the process, then you implement them as cron jobs and skills. “The harness doesn’t really matter as much as just the skills.” The thing most teams never build: agreement-rate measurement between humans and the agent — candidates a recruiter passed that Jim rejected, standouts Jim flagged that no recruiter reached out to. Why the PM playbook changed Pedro’s tweet: the PM playbook of writing PRDs and aligning stakeholders is dead. His argument: the quality of a product org is how fast it compresses signal → ship, and AI eats most of the ingest-and-digest half. “PM as mini CEO” is out at Brex; engineering and design craft sit on equal footing with PM. The roles are conflating: PMs write code, designers build, engineers make calls that used to be PM territory. What PMs do now: ingest six months of Gong calls, unearth insights from unstructured data, validate with cheap prototypes — so much conviction accumulates that you ship the right thing first. Because building is cheap, discernment matters more than ever: pick the one true lever, minimize surface area, go deep on one thing rather than shipping five or ten. Half the job is reviewing work The case-law model: leadership review is the Supreme Court that sets the standard, then reviews cascade down to the IC building the thing. Two review types at Brex: problem alignment (which problem is worth solving) and solution alignment (is this the right fix). Brex keeps a repo of PM skills so a review arrives already digested — humans then spend time only on what is idiosyncratic to that problem. Standing questions: what were the alternatives, what is the actual bottleneck, how does this create more throughput on it. “We manage the work, not the people, and we manage people around the work.” Hiring: the interview is a builder loop — you have to build something with AI proficiency. Open source presence is a positive signal; its absence is not a negative. CrabTrap: securing agents at the network boundary Brex open-sourced CrabTrap, an LLM-as-a-judge HTTP/HTTPS proxy that sits around the agent and controls all its network traffic (brexhq/CrabTrap). Reasoning: if the agent can run code — and it should — it can always cross a network boundary. Constraining tools alone doesn’t hold. Every request hits static rules first (URL prefix/exact/glob → immediate decision, no LLM call); anything not matched goes through a policy prompt evaluated by an LLM acting as judge. You don’t have to write the policy: CrabTrap replays observed traffic, summarizes it with a model, and proposes the ruleset (~30 minutes to run). This is semantic, not allow-list: it can research candidates on GitHub but not push to repos; it can’t delete repos or change Okta configs. Granola was blocked — the agent could reason about why. Cost of a block: about 1,000 tokens in, 104 out, ~2 seconds on Sonnet. Audit-trail mode shows every request live. Magpie: the token bill is coming Brex built Magpie for AI cost visibility, split into three pillars: corporate AI (internal productivity), operational AI (serving customers, automating processes), product AI (shipped features). The point is granularity by caller: transaction tagging costs $0.10 a call ($28k/month), disputes ~$2 a call — a 240x spread. Plotting cost-per-call against call volume shows what actually changed. You can see which harnesses dominate (Claude Code first, then Codex, Cursor), cost per customer, per employee, and who is tokenmaxxing. Pedro’s sequencing advice: let people go deep first, optimize later. Brex gives unlimited token budgets with a few caps, then surgically attacks low-ROI usage — caching, or moving work to cheaper models. The reconciliation angle: Brex is the card and the rail, so usage data can be matched against the dollars that actually clear the bank account. This becomes a Brex product. Pricing is shifting from per-seat to per-token, including products that merely wrap tokens (Cursor) — so even an all-Anthropic shop has token spend scattered everywhere. We are still early — the 2,500-box chart Pedro’s favorite slide: the world in 2,500 boxes, each dot 3.2 million people. Red = never used AI; green = free chat; orange = paying $20/month; and a tiny box = people using agents effectively. His thesis on jobs: role conflation, everybody a builder, a market for 10x employees — raise the floor and the ceiling at once, with the ceiling rising far more. Dispersion in productivity is normal (10x engineers); AI widens it while lifting the baseline. Why shouldn’t a company founded today be you, a big token budget, and agents? Autopilot: the harness that runs his life Pedro’s personal system, Autopilot, is an OpenClaw harness with two building blocks — people and programs — plus signals: Slack messages, email, meeting notes. A signal collector runs on a schedule; an aggregate job then goes through every person and program and updates the markdown files (status, blockers, risks, action items). Drafts appear automatically from signal injection: a Slack DM asking about a conference became a task to decide on it, with the DM attached as context. Everything is markdown files committed to a repo, with scripts generating searchable and editable UIs; an interaction log traces each task back to the signal that created it. On OpenClaw reliability: “an acquired taste… definitely takes a bunch of time to do it well,” but worth it. Selling the work, not the tool The SaaS subscription era is blurring: the currency changed to tokens, but you’re still paid for solving a problem — you now charge a markup on tokens instead of a seat. The bar for a good AI product is higher, because you can sell the work itself rather than the software. “The reality of all good AI products is they’re all the same thing: an agentic loop and a measure of tools.” “You expose the tools to the model, you run it on an agentic loop, and just let it do its thing.” ...
Cole Medin’s full walkthrough of standing up his open-source “software factory” — the autonomous issue-in, merged-PR-out harness he’s been building this year — on a remote VPS, with Codex running GPT-6 Astra as the worker model (16 min). On the AGI talk Jensen Huang reportedly declared AGI achieved because of Astra. Medin’s response: don’t buy the hype — and then concedes he’s starting to buy into it, just not the AGI part On benchmarks Astra and Fable 5.1 look equivalent. After a week of head-to-head testing, he says Astra comes out on top a majority of the time His specific claim: it’s the first model that “gets” him — far less time spent communicating intent, and notably fewer weird assumptions than Opus 5 made over the past few months Still “has its moments.” That’s his evidence that we are not at AGI, whatever the benchmarks show What a software factory actually is The pitch: a fully autonomous harness where the PRD goes in and shipped code comes out — he also calls it the “dark factory” Earlier this year the idea was unrealistic; the models and the harnesses weren’t reliable enough. With Astra he says he can see the light at the end of the tunnel Honest framing throughout: not close to 100% reliable for arbitrary code, and he calls the project early alpha while he keeps refining it Operating loop, in his description: Input is always an issue (a PRD split into issues, or issues you write directly) A triage workflow reads the issue against the factory’s rules and mission context, decides accept or reject, and ranks by priority It runs the work through to a pull request that is fully reviewed Most of the time it merges automatically; it escalates to a human only when it has to He’s been using it on his own projects — Archon (the open-source harness underneath), games, whatever he’s building that week The deploy, which is mostly prompts Two supported starting points: a brand-new repo where you only have a PRD, or bolting the factory onto an existing codebase. His demo uses a small existing link-shortener app because it means less initial setup The whole setup is: give your coding agent the cheat-sheet URL from the repo and say “help me set this up.” It interviews you to establish the factory’s core context files, then walks the deployment He deploys with Codex on GPT-6 Astra, though he says Claude Code and Codex are the two he’s tested heavily and either works Hostinger’s MCP plugin lets the agent manage the VPS directly (this video is sponsored by Hostinger; the real requirement is just an Ubuntu box in the cloud). He installs the plugin user-scoped, authenticates in the browser, then asks it to list his instances as a smoke test The plugin deliberately can’t create instances — he doesn’t want an agent holding that power — so you spin up the VM yourself and hand the agent the VM ID and public IP The agent then rips through SSH access, firewall rules, and installing the factory’s dependencies Two steps stay manual, because you shouldn’t hand over credentials: GitHub auth and OpenAI/Codex login. Both use device-code flows you complete in a browser Gotcha worth copying: enable device code authorization for Codex in ChatGPT settings → Security and login, or remote auth gets needlessly painful Sanity check on the box: codex exec a hello to confirm the model answers before continuing Tell the setup agent “I’m done” and it installs the factory plus Archon, confirms Codex is live, and installs your target repo Proving it end to end The cheat sheet doesn’t stop at install — it drives a first test issue, created through the GitHub CLI in your repo He takes the agent’s own recommendation for the issue rather than inventing one The factory triages it (marked “archon ready”), runs the workflows, and produces a pull request The test he ran was trivial on purpose — the point is proving the trigger path and the workflows work with the coding-agent auth on the remote machine After the first PR validates, you close the setup session. The factory keeps running 24/7 on the box, accepting any GitHub issue you file The agent can even update DNS records through the same plugin, if your domain is already there Caveats he states plainly Still early alpha, still not reliable for everything, and he’d be the first to say so The guide is intentionally platform-flexible, which means parts are high-level: expect the agent to hit snags and hand you commands to run, especially around authentication He’s promising more content on how the internals work and how to make the whole thing token-efficient — that’s the part this video skips “We literally go from issue all the way to validated and merged code.” — Cole Medin ...
Hamel Husain and a co-host spend an hour live-testing claude plugin eval — the plugin-evaluation feature Anthropic shipped for Claude Code (announced from the Claude Devs X account) — by pointing it at a plain-writing/tone skill. They never get a useful eval out of it, and the diagnosis of why is the interesting part. ...
The Next New Thing’s Andrew Warner and Adam Brakhane run through this week’s GitHub roundup, and it lands heavily on skills: small packages that tell a coding agent how to think before it writes anything. Andrew opens by calling it a skills-heavy week, and Adam spends the episode arguing that most of these skills package something you could just say out loud. Then OpenAI deprecates its skills repo mid-roundup, which kind of makes Adam’s point for him. ...
Rob Kaminski — co-founder of Fletch PMM, the consultancy that has run positioning work for roughly 600 B2B startups including Notion, GitLab and Superhuman — does a live teardown with host Angelina on TwoSetAI (1h02). The subject is a real AI agent startup: AgensFlow, Nicole Königstein’s coordination layer that sits above frameworks like LangChain and LlamaIndex and learns which agent and model to use for each task. ...
Cole Medin on the security gap in AI coding workflows (17 min, partnered with Sonar; Archon, the harness builder he uses, is his own open-source project). His premise: agents are already better and faster than humans at a lot of code — but security is where they fall short far more than reusability, documentation or performance, and almost nobody is building for it. ...