Human in the Loop: Two Ex-Microsoft Builders on AI

Chris Shanku
  • 0.0 (0)
  • Technology
  • Updated Weekly

Chris Shanku and Sash Mohapatra - two ex-Microsoft builders - demo one real AI tool per episode, live and unedited. Real people. Real tools. No retakes.

  1. 3d ago

    Stop Running Agents Without Shared Memory

    Sash drives this one, and he does not start with the tool. He starts with the failure. He was running Claude Code across two Macs, so his first move was the obvious one: put his intelligence folder of Markdown context files in a git repo and sync it to both machines. That worked until it did not. Every session ended with a push he had to remember, and every session started with a pull he had to remember, and forgetting either one meant an agent was working from yesterday's truth. Then he added a Hermes agent on a Hostinger VPS, then revived an old OpenClaw agent on another one, and suddenly there were four agents and no shared understanding between them. The fix was not better discipline, it was a different layer. Sash looked at Obsidian vaults and Karpathy's LLM Wiki before landing on gbrain, the open-source project from Y Combinator's Garry Tan, which Tan describes on the repo as "the production brain behind my OpenClaw and Hermes deployments." gbrain sits on top of the git repo as a semantic layer: it chunks the Markdown, indexes it into Postgres with pgvector, and exposes an MCP endpoint agents can query. The git repo stays the durable source of truth and stays human-editable. What changes is retrieval. Instead of loading CLAUDE.md plus status.md plus decisions.md into every context window, an agent asks gbrain for the chunk that matches the task and pulls more only if it needs it. That is the moment Chris stops the demo to repeat it back, because just-in-time context is the part that actually changes a working day. The second half is the part most people skip, which is access. Sash runs three agents against one brain with deliberately different permissions: Nuro is the only thing with write access to the git repo, Kyra has full read/write on gbrain but read-only on the repo, and Blaze can only read the course-related slice and cannot see client work at all. His framing is that this is not novel, it is just how you build an IT team - help desk does not get admin on the database, engineers do. Sash closes with the smallest version worth doing: if you run one agent and nothing always-on, hand the gbrain repo to Claude Code or Codex, point it at the files you already work in, and describe the thing your agent keeps forgetting. Key takeaways: Put the durable source of truth in a git repo of plain Markdown you can still open and edit by hand, then treat gbrain as an index on top of it rather than a replacement for it.Stop loading whole context files into every session. Wire a hook so the agent queries the brain for the chunk the task needs, and let it pull the full page only when the chunk is not enough.Separate the three things an agentic system needs: context lives in the brain, canonical instructions stay in CLAUDE.md or AGENTS.md, and tools are wired per agent.Decide read versus write per agent before you connect the second one. Give exactly one agent write access to the source repo and scope every other agent to the slice it actually needs.If you are starting cold, do not install it by hand. Give the gbrain repo URL to Claude Code or Codex, tell it what your agent keeps forgetting, and let it work out the setup and the sync job.Watch the video version on YouTube · Full show notes, transcript + takeaways

  2. Aug 19

    AI for Enterprise: Stop Starting with AI

    Kevin Johnson - KJ - leads the AI and automation practice at Pinnacle Financial Partners, a bank that closed a merger of equals in January between two institutions north of $60 billion each. He has been doing this hand-to-hand since ChatGPT launched in November 2022. Chris and Sash brought him on to answer the question every enterprise team is quietly stuck on: what actually has to be true before generative AI works in production. His answer is that almost none of it is about AI. Pinnacle's ability to move fast on generative AI traces back five or six years, to a modernization journey that ripped workloads out of a traditional data center, rebuilt the data and networking and security layers, and reorganized teams into pods around the new work. That unglamorous foundation is the reason the AI work has anywhere to land. KJ's sharper point is that the same modernization has to happen to people, and that half gets far less attention: reinventing yourself as a contributor after two decades of doing the job one way is hard, and the organizations that win will be the ones that move fluidly through that change. As he put it, that is as human as it gets. The back half is about discipline. Value in generative AI is genuinely hard to measure, and KJ is blunt that if you are not already measuring your KPIs well, layering AI on top makes measurement harder rather than easier. He and Sash land in the same place on tooling: automation and AI are two different tools in the toolbox, deterministic solutions never actually died, and you do not need a cannon or a bazooka for a mosquito. Chris connects that back to skills as the middle ground - a way to get an agent to follow the same process every time, even when the wording varies - and notes how often his skills end up becoming automations. The advice KJ would give his 2023 self is not technical at all: get real leadership buy-in, make continuous innovation part of the DNA, and meet the organization where it actually is. Key takeaways: Do the modernization work first. Cloud, data, networking, and security foundations are what make a generative AI program possible, and skipping them is the same mistake teams made pushing BI projects ahead of data cleansing.Budget for team member modernization, not just technology modernization. The hard part is people reinventing how they work, and that is a change management problem rather than a tech problem.Fix your measurement before you add AI. If your KPIs are not instrumented well today, proving generative AI value on top of them gets harder, not easier.Match the tool to the problem. Deterministic automation and RPA still solve plenty of things more cheaply and more reliably than a frontier model, and orchestration plus observability matter more than horsepower.Use skills to put determinism back into agent work. Define the pattern once so the process runs the same way every time, and expect the good ones to graduate into automations.Ship tools that make someone's day measurably easier and adoption becomes organic. Finding an answer in ten seconds instead of two and a half minutes is the kind of win a frontline team feels immediately.Watch the video version on YouTube · Full show notes, transcript + takeaways

  3. Aug 14 ·  Bonus

    Here Comes Watermarking!

    Anthropic started watermarking every piece of text Claude produces, worldwide, three days before this episode and Sash and Chris jumped on an unplanned morning show to work through what it actually does. The watermark has two parts: an invisible mark woven into the text itself, and C2PA metadata (the same provenance standard Microsoft and Google use) attached to generated artifacts like images and documents. It is not opt-in - Anthropic is required to ship it everywhere under Article 50 of the EU AI Act, the same transparency requirement that is starting to do for AI content what GDPR did for data. The conversation kept circling one unresolved question: the mark does not distinguish between text Claude wrote from scratch and text a human wrote and ran through Claude for a light edit, a fact-check, or a translation. Both cases get flagged the same way. Sash pointed out the mark also strips cleanly the moment you run the output through a second model to rephrase it, which means the ambiguity cuts both directions - it can wrongly implicate a human editor, and it is trivially defeated by anyone trying to hide real AI generation. Anthropic says a public text-detection API is coming, which is the piece that would actually make this enforceable rather than theoretical. Both hosts landed on the same personal practice out of this: get in front of it and disclose AI use yourself instead of waiting to be caught by a detector. Sash already footnotes every newsletter issue as written with Claude Code; Chris said the episode convinced him to start doing the same with his own AI-generated text going forward. Key takeaways: If you publish AI-assisted writing, disclose it yourself in the footer or byline rather than waiting for a detection tool to flag it later - both hosts already do this and treat it as the safest default under the new mark.Don't assume a light-touch AI pass (grammar check, fact-check, translation) is exempt from the watermark - the mark applies to anything that passed through Claude, not just fully AI-authored text, so budget for that before you rely on it staying invisible.Know that the watermark is currently defeated by running Claude's output through a second model to rephrase it - if you or your org are planning to lean on detection for policy or trust decisions, that gap matters today.Watch for Anthropic's promised public text-detection API - that is the step that turns "a mark exists" into "someone can actually check for it," and it is not live yet.Watch the video version on YouTube · Full show notes, transcript + takeaways

  4. Aug 12

    3 Lessons from Blowing up my AgentOS

    Rene Charbonneau returns three months after EP.009 to report on the promise he made on his way out the door: refactor his AgentOS. He didn't refactor it. He blew it up. What started as a two-week attempt to untangle a ball of yarn he and Claude had built together turned into a full rebuild once he realized the old assumptions (a single hand-built harness, a builder-first workflow, memory scattered across machines) no longer held. The teardown included literally retiring his always-on agent, Jim, mid-project, and returning to rebuild the connection layer (SSH, approvals, skills) from scratch on a fresh machine profile. The conversation centers on what Rene replaced the old setup with: less building, more configuring; a shift from Claude Code alone to a Hermes/OpenClaw split, backed by an LLM Wiki-style agentic memory layer (Y Combinator's G-brain) synced read-only across multiple machines so every agent works from one source of truth. He also dropped his Claude subscription from $200 to $20/month, using Hermes and OpenClaw against OpenAI instead, partly for cost, partly because routing agent traffic that way violates Anthropic's terms of use, and partly because watching Fable get geo-restricted in Canada made vendor lock-in feel like a real risk rather than a hypothetical one. The back half zooms out to where this goes for non-technical users: Microsoft Scout and rocket-branded turnkey "agent PCs" (Auto, VioLoop) as the consumer packaging layer for the same OpenClaw/Hermes stack the hosts run by hand, POKE as a prepackaged agent you negotiate with over iMessage, and a BYOA ("bring your own agent") framing for how enterprise hardware procurement might look in five years. Chris teases next week's guest, Kevin Johnson of Pinnacle Financial, who will cover the enterprise version of this same journey: foundation-first before AI, not AI bolted onto ungoverned data. Key takeaways: If something about your agent setup keeps nagging at you ("is this friction normal?"), treat it as signal - spin up a parallel instance with a stock model and no custom harness, and compare results before assuming your custom setup is still worth its context budget.When you retire an agent or harness, don't delete its memory - archive it as a tombstoned "V1," ask it what it wants carried forward into "V2," and start clean from there.A shared, synced memory layer (wiki-style, read-only mirrored across machines) matters more than which harness you pick - organize it so every agent can actually find what it needs, not just access it.Watch for vendor lock-in risk directly, not just cost: geographic restrictions (Fable's brief US-only limitation) and terms-of-use constraints (routing autonomous-harness traffic through Anthropic directly) are real reasons to keep a second provider live.Scope what an always-on agent can do to itself - Rene's agent needed a human to step in and grant runtime access precisely because it was walled off from troubleshooting its own failures, which is what kept a bad actor or prompt injection from having free rein.Watch the video version on YouTube · Full show notes, transcript + takeaways

  5. Aug 5

    Stop Doing Video Production Manually

    Chris opened the show with roughly thirty seconds of himself that was not himself. The voice, the face, and the script were all generated, and the reveal was the point: the pipeline that produced that cold open is the same one that produces the trailer, the highlight reel, and the product demos. He then kicked off two real jobs live on stream, the EP.020 trailer and an iLink product demo, and let them run in the background while he and Sash talked through how the machinery works. The shape of the pipeline is the same everywhere Chris uses it. A research skill finds material inside a defined niche. A script skill drafts it, grounded in enough of his own writing that the draft comes back sounding like him rather than like a model. A critic agent checks the draft against the storytelling rules he cares about and sends it to Telegram for approval. Only after that gate does a produce skill start spending money, calling ElevenLabs for voice and rendering images. A release skill then schedules the finished video to YouTube against a defined cadence, cuts the Shorts, builds the Facebook and Instagram posts and reels, and pushes the audio to the podcast RSS feed. Nothing in the orchestrating skills holds logic; they call subskills, so a single stage can be changed without touching the rest. Sash's framing for all of it was deconstruct and delegate. His argument was that a complex multi-step task should never be one pipeline, and that the three things that decide whether long-running agent work succeeds are context, instructions, and tools, set up per piece rather than once globally. He tied it back to spec-driven development: nobody writes production software without planning the user journey and the features first, and agent pipelines deserve the same treatment. The honest moment came at 29:21, when the product demo Chris had kicked off did not finish end to end despite testing clean three times beforehand. Nobody cut away. He showed a previously rendered demo instead and moved on, which is roughly the whole thesis of the show. Key takeaways: Put the human gate where the money starts. Chris approves the script on Telegram before the produce skill runs, because everything upstream of that point is free and everything downstream costs real API spend.Keep orchestrating skills free of logic. Have them call subskills instead, so you can change the voice step without editing the whole pipeline or hoping the model does not break something adjacent.Track cost per minute of finished video as your KPI. Chris lands around 50 cents per minute for a story and a little over a dollar per minute for a product demo, counting images, voice, digital twin, and music.Deconstruct before you automate. Write down the whole pipeline from research to release first, then turn each phase into its own skill with its own context, instructions, and tools.Taste is not a prompt, it is time. The way to get output that matches your creative vision is to iterate and review repeatedly, and to define what good looks like in words the machine can act on.Watch the video version on YouTube · Full show notes, transcript + takeaways

  6. Jul 31

    The Harness Outweighs The Model

    Sash drove this one, and the framing he opened with does most of the work: a model is a faucet of intelligence, but a faucet on its own does not plumb a house. An agent is a model plus a harness, and the harness is the environment the model actually runs inside. It is the app. The chat window is a harness. Claude Code is a harness. Hermes and OpenClaw are harnesses. What the harness supplies is everything the model cannot supply itself: tools, memory, skills, and a scheduler. Take those away and an agent collapses back into a chatbot. The useful part of the episode is the ladder Sash laid out. Chat sits at the bottom, the simplest harness there is, with almost nothing to break. A session-based agent harness like Claude Code or Codex sits above it, capable of pulling tools, running skills, spawning subagents, and looping. Always-on harnesses like Hermes and OpenClaw sit at the top, waking on a schedule, reading a task list, and working without a human in the session. The catch is that fragility climbs right alongside capability. Chris and Sash both hit it: OpenClaw dying silently on background tasks with no signal until you go turn on traceability, and Chris discovering he could not take a clean update after forking Hermes, because customization is exactly what makes an always-on harness yours and exactly what makes it yours to maintain. The number worth carrying out of the episode is 8x. Sash cited a controlled study showing the right harness multiplied agent capability roughly eightfold with the model held constant. That reframes where the leverage is. Chris landed the analogy that ties it together: the harness is to agents what retrieval augmented generation was to GPT three years ago, the surrounding machinery that turns a generic model into something grounded in your context and useful on your problems. Both hosts expect enterprises to start building their own harnesses, partly to run cheaper open-source models and partly to hedge the cost and geopolitical exposure of depending on frontier providers. Harness engineering is already showing up as a taught discipline, which is usually the sign that something has stopped being a niche. The show also changes format starting next week: YouTube instead of LinkedIn, Wednesdays at 2 PM ET instead of Fridays, and 25 minutes instead of 15. Key takeaways: When an agent underperforms, change the harness before you change the model. A controlled study put the capability difference at roughly 8x with the model held constant, so the scaffolding is usually where the leverage actually is.Pick the lowest rung of the ladder that does your job. Chat, session-based agent, always-on. Capability and fragility climb together, so buying capability you do not need means buying failure modes you did not want.Instrument always-on agents for silent failure before you rely on them. OpenClaw dying quietly on a background task is invisible until you deliberately enable traceability, and an agent that fails without telling you is worse than one that does not run.If you fork an always-on harness, you own its updates. Chris forked Hermes, customized it, and then could not take a safe update. That is the real price of the control, and it is not payable in AI time.If you are at a company weighing frontier model dependency, start scoping what your own harness would look like now. The argument for it is cost and concentration risk, not capability, and the tooling and teaching around harness engineering is arriving quickly.Watch the video version on YouTube · Full show notes, transcript + takeaways

  7. Jul 24

    My Agent Writes My Emails. A Review Gate Decides If They Go.

    Chris drove this one solo-demo style, no guest, closing out the AI-for-enterprise arc for now to make room for other topics. He walked through TheGrove, his personal agent system, and specifically the Follow-up Engine: an agent that reads meeting notes and call transcripts and drafts the follow-up email in his own voice. The demo used a sanitized mock transcript (a made-up company, "Harborline Financial") - the agent identified attendees, pulled out three to five commitments, let Chris reassign a miscategorized commitment to the right person, and then drafted a follow-up email trained on hundreds of his real emails and Teams messages for tone, format, and sentence length. The core mechanic: the agent only ever produces a draft. It lands in a review gate before anything sends, and the system logs the diff between what the agent wrote and what Chris actually sent - word changes, tone changes, punctuation - so it keeps learning from his edits over time. Every Friday he gets a "Follow-up Friday" digest on Telegram: how many emails sent, average edit rate, how many went out verbatim. The goal isn't a smarter agent, it's fewer edits, earned one draft at a time. Chris also showed the daily morning digest (calendar, blind-spot detection between work and personal calendars, overdue commitments) and the one-click "action debrief" summary send, which has a safety-mode toggle so it can't accidentally reach real customers. They closed by teasing next week's episode: the harness (not the model) is often what actually determines agent capability, and they're breaking down the spectrum from chat to agentic to always-on autonomous systems. Key takeaways: Constrain what the agent is allowed to generate (one paragraph, not a whole email) so the review step stays fast and the failure mode stays small.Log the diff between agent draft and human-sent output. That diff is the training signal for reducing future edits, not just an audit trail.A "blind spot" check (comparing personal vs. work calendars) is a simple, high-value use of an always-on agent that has nothing to do with generation quality.Feed an AI writing tool your own transcripts and sent-item history if you want it to actually sound like you, not a generic assistant voice.A safety-mode toggle on any agent action that can reach a real customer is a cheap insurance policy against an expensive mistake.Watch the video version on YouTube · Full show notes, transcript + takeaways

  8. Jul 17

    AI for Enterprise: Institutional Knowledge

    This is episode four of the AI for Enterprise series and the most practical one yet. Sash walks through what Claude skills actually are - a folder with a skill.md file written in plain English, plus any reference files, templates, or scripts the workflow needs. The standard was released by Anthropic in November 2025, and within two months 16 other vendors adopted it, including Microsoft (it shows up as skill.md in M365 Copilot Agent Builder). That portability is the point: a skill you write for Claude today can move to Replit or Cursor unchanged. The bulk of the conversation is about why skills beat agents for most enterprise use cases. Sash's rule of thumb is direct: only reach for an agent when a skill would overflow the context window and lose quality. Otherwise a skill running inside Claude chat or Claude Cowork delivers consistent, predictable output at predictable token cost - which is exactly what you need when ten SDRs or twenty marketers are all hitting the same process. He showed his own skill library live: a simple create-doc skill with just a skill.md, versus a multi-agentic deep-research skill with four reference files providing process context. Chris tied it to the enterprise governance angle the series has been building toward. Skills are a pro-code-to-low-code bridge: your AI architects build and vet the skills, the Centre of Excellence promotes them, and frontline workers get approved, consistent AI behavior at scale. Chris also confirmed this entire show - the episode pipeline, website updates, everything - runs on orchestrating skills that call other skills. The episode closes with a tease: the series is getting a fifth week, and they're bringing in a guest from the enterprise world. Key takeaways: A skill is a folder with a skill.md (plain-English instructions) plus optional reference files, templates, or scripts. No code required to create one.Use an agent only when a skill would overflow the context window. For everything else - outreach sequences, briefing, battle cards, brand copy - a skill is cleaner and cheaper.Claude Teams and Claude Enterprise let you provision skills at three tiers: owner-only, team-level, or org-wide. Marketing gets marketing skills. Sales gets sales skills. You can centrally update and evaluate them.Let Claude build the skill for you. Run the process in chat, iterate until the output is right, then ask Claude to package it into a skill. It reads the conversation history and produces the skill folder.Skills created in Claude chat and Claude Cowork are interchangeable. Skills created in Claude Code use a different packaging mechanism and will not appear in Cowork - you may need to recreate them.Watch the video version on YouTube · Full show notes, transcript + takeaways

See All (22)

About

Chris Shanku and Sash Mohapatra - two ex-Microsoft builders - demo one real AI tool per episode, live and unedited. Real people. Real tools. No retakes.

Information