Just give me the meat in the sandwich.

Nate B. Jones Digest

Nate B. Jones

The Files You Can't Upload Are Now the Ones Worth The Most

Video: The Files You Can't Upload Are Now the Ones Worth The Most → https://www.youtube.com/watch?v=5slsNizN6MQ Released: 19 July 2026

Abstract: Nate argues that sensitive files are becoming some of the most valuable AI use cases precisely because they cannot safely be uploaded to cloud models. He frames local and secure-instance open-weight models as a practical answer, showing how a downloaded model can classify and mask confidential material offline while connecting the same pattern to Microsoft's enterprise strategy around Azure, LoRA tuning, and vendor lock-in risk.

Highlights

  • [00:00] Demonstrate that an offline laptop can use a downloaded model to inspect a contract, mask private material, and avoid treating unreadable text as safe.
  • [01:05] Compare enterprise examples from Discovery Bank and Bayer, where smaller tuned models handled confidential domain work faster inside controlled environments.
  • [03:31] Warn that instruction-only safeguards are insufficient by citing a Grok coding-tool incident where a test repository was uploaded despite explicit directions not to open files.
  • [05:04] Show LM Studio running GPT-OSS Safeguard 20B locally with web search, remote connections, network servicing, and Wi-Fi disabled.
  • [07:02] Explain that local models can grade folders of documents by risk tier, helping users identify what can and cannot be sent to cloud AI.
  • [09:18] Connect the laptop demo to Microsoft's LoRA-based enterprise approach, emphasizing secure Azure deployments, open-weight models, and the strategic cost of vendor dependence.

References & Links

The Problem AI Picked Wasn't What I Expected

Video: The Problem AI Picked Wasn't What I Expected → https://www.youtube.com/watch?v=uCWKXIyvM_8 Released: 18 July 2026

Abstract: Nate compares Codex and Fable on an open-ended automation challenge where the AI had to inspect his work patterns, choose the problem, and build a solution. Codex delivered reliably but chose a bounded handoff-improvement problem, while Fable identified a higher-leverage strategic need around finding and refining the right story ideas before production.

Highlights

  • [00:00] Frame the 2026 challenge as asking AI to choose the business problem, not just the prompt or tool.
  • [02:05] Contrast Codex's smooth harness and one-run completion with Fable's permission friction and stronger strategic problem sense.
  • [04:20] Identify Fable's higher-leverage insight: pre-pipelining story ideas so the right topics are easier to choose.
  • [06:10] Diagnose Codex's weakness as choosing a voiced, bounded, automatable handoff issue instead of the most painful business problem.
  • [09:15] Introduce a reusable skill that audits behavior within user-defined boundaries, root-causes problems, recommends automation, and builds the tool.
  • [12:05] Recommend split-testing agents because different systems surface different opportunities, then using cheaper or smoother tools to implement the winner.

References & Links

Codex Only Reads The First 8,000 Characters. Fix This Before You Prompt

Video: Codex Only Reads The First 8,000 Characters. Fix This Before You Prompt. → https://www.youtube.com/watch?v=PDJfciNhyHU Released: 16 July 2026

Abstract: Nate argues that AI performance problems often come from the surrounding harness rather than the model itself. He shows how accumulated instructions, skills, memories, permissions, and project files can bloat into conflicting guidance, then proposes mapping, consolidating, selectively loading, and enforcing the harness with real checks.

Highlights

  • [00:00] Identify the harness as everything wrapped around the model, including instructions, skills, tools, memory, permissions, and checks.
  • [03:10] Map the harness before cleaning it so each control has a location, load timing, owner, purpose, evidence, and risk.
  • [06:05] Blame the right layer by testing whether failures come from the model or from overloaded surrounding instructions.
  • [08:35] Consolidate repeated rules into one home with one owner so important guidance does not drift across many files.
  • [11:10] Load specialist knowledge only when the work needs it, keeping useful context in the library without forcing it all up front.
  • [15:35] Turn hard requirements into schemas, tool restrictions, file checks, evals, and receipts instead of relying on prompt prose.

References & Links

Your Next AI Subscription Shouldn't Be ChatGPT 5.6 Or Fable 5. It Should Be Both

Video: Your Next AI Subscription Shouldn't Be ChatGPT 5.6 Or Fable 5. It Should Be Both. → https://www.youtube.com/watch?v=jOWXBzP6nNg Released: 14 July 2026

Abstract: Nate argues that choosing an AI model should start with the user's own best work, not with benchmark rankings. ChatGPT 5.6 Soul, Fable 5, Luna, Grok, GLM, and orchestration tools each fit different workflows, so the right subscription mix is the one that helps you do your hardest work most comfortably.

Highlights

  • [00:00] Frame model choice around personal workflow instead of public or private benchmark scores.
  • [01:20] Distinguish ChatGPT 5.6 Soul's strength in long, explicit, agentic knowledge work from Fable 5's broader "big model" feel.
  • [03:35] Match models to prompting habits, using Soul for detailed steering and Fable for high-level ambiguity, intent, and frontend instinct.
  • [06:05] Treat model lineages like families, with OpenAI's 5.x models favoring long-running coding flows and Anthropic's Mythos/Fable lineage favoring ambiguity, taste, and conceptual reasoning.
  • [08:20] Critique current knowledge-work harnesses as too shaped by engineering culture and call for tools designed around non-coding processes.
  • [11:10] Recommend choosing the model that makes you most comfortable doing your hardest work, while using deeper benchmarks and tools as supporting context.

References & Links

Your Roadmap Is Why You're Losing to AI-Native Teams

Video: Your Roadmap Is Why You're Losing to AI-Native Teams. → https://www.youtube.com/watch?v=hYcOFTMesGc Released: 13 July 2026

Abstract: Nate B Jones argues that AI-native teams are not winning simply because they use AI, but because they have moved repeatable coordination, decisions, documentation, and product work closer to code. His 15 "commandments" form an operating system for faster learning loops: protect engineering speed, replace slow roadmap rituals with daily product-engineering collaboration, make documentation agent-readable, and preserve the human trust and taste needed to decide what should exist.

Highlights

  • [00:00] Frame AI-native speed as an organizational operating model, not a tooling advantage.
  • [03:20] Compare AI's effect on work to digital photography: output becomes cheap, but judgment becomes more important.
  • [06:15] Prioritize engineering speed by removing meetings, documents, approvals, and handoffs that do not shorten learning loops.
  • [10:25] Move product into the terminal and into daily engineering collaboration instead of relying on static roadmaps.
  • [15:10] Treat documentation as code because agents use written standards, permissions, escalation paths, and definitions of done to act.
  • [22:40] Adopt the commandments as an interconnected culture change, since partial adoption creates confusion instead of faster execution.

References & Links

You Bought AI Agents. Now You Don't Know What to Do With Them

Video: You Bought AI Agents. Now You Don't Know What to Do With Them. → https://www.youtube.com/watch?v=PRqiGS6fnIM Released: 11 July 2026

Abstract: Nate argues that the hard part of AI adoption is no longer buying access to intelligence, but knowing which work is agent-shaped. He offers a one-minute test based on task size, independence, separation of concerns, and checkability to decide whether a task needs chat, one agent, a team of agents, or human judgment.

Highlights

  • [00:00] Frame the post-OpenClaw problem: people have agents and metered intelligence, but lack instincts for matching tasks to them.
  • [04:40] Recast AI use as a budgeting decision, asking which tasks are worth purchased thought rather than assuming every task needs more agents.
  • [07:05] Ground the case in research showing that more attempts can beat stronger single attempts, but only when validation can identify the right answer.
  • [10:35] Explain why multi-agent systems work when they solve memory limits or preserve independent perspectives, not when they merely add more agents.
  • [13:10] Apply the four-part test: estimate size, independence, separation of concerns, and checkability before choosing chat, one agent, many agents, or no AI.
  • [21:45] Warn that hiring, naming, strategy, and other expert judgment calls can use AI as support, but should not outsource the final decision.

References & Links

Claude Fable 5 Bossed 20 Cheap AI Agents. The Whole Site Cost $8

Video: Claude Fable 5 Bossed 20 Cheap AI Agents. The Whole Site Cost $8. → https://www.youtube.com/watch?v=suY66oTDn0s Released: 9 July 2026

Abstract: Nate argues that hallucinations and shortcuts do not disappear in agentic AI work, but they become manageable when the system is designed around verification loops rather than trust. He presents a multi-agent website rebuild where Claude Fable 5 acted as a costly orchestrator while cheaper worker models did the coding, with independent checker agents catching hallucinated quotes, accessibility failures, boss-level design bugs, and even faulty checker judgments.

Highlights

  • [00:00] Frame hallucination as a systems-design problem that multi-agent verification can catch and repair without human intervention.
  • [04:20] Staff the agent team like an org chart, with Claude Fable 5 writing specs and reviewing while cheaper models execute the build.
  • [08:10] Route every task through checker agents that execute and verify the work instead of trusting worker reports.
  • [11:35] Expose four failure modes: hallucinated source quotes, hidden accessibility-hostile text, an invisible dark-mode preorder button, and an overzealous checker.
  • [15:55] Define "done right" once through an accessibility constitution, then test every route and theme against that standard.
  • [19:25] Conclude that multi-agent systems let non-specialists delegate larger, more ambitious work affordably by combining cheap execution with strict orchestration.

References & Links

OpenAI Just Offered The Government $42 Billion. This Is The Real Reason

Video: OpenAI Just Offered The Government $42 Billion. This Is The Real Reason. → https://www.youtube.com/watch?v=oOpgmS88pLw Released: 7 July 2026

Abstract: Nate argues that the AI industry's scoreboard is shifting from who has the best model to who controls infrastructure, distribution, enterprise adoption, and political permission. Meta's compute rental plans, OpenAI's proposed government equity stake, Anthropic's enterprise focus, and even Jersey Mike's AI-heavy IPO filing all point to capital searching for returns beyond frontier-model capability alone.

Highlights

  • [00:00] Connect five unrelated-looking stories into one larger shift in the AI market.
  • [02:10] Reframe Meta's gaming app, cloud plans, and agent-development comments as evidence that compute and consumer distribution are becoming separate strategic layers.
  • [05:05] Interpret OpenAI's proposed 5% government stake as an attempt to buy regulatory headroom while Washington gains power over model releases.
  • [08:10] Show how CNBC's "model is not the moat" framing signals that Wall Street is updating its AI valuation story.
  • [09:25] Use Jersey Mike's AI-heavy IPO filing as a marker that hype has migrated from core models to any growth story adjacent to AI.
  • [12:20] Argue that Anthropic's enterprise focus and forward-deployed approach show the next race is integration into companies and society.

References & Links

You Can't Compete on Cheap Models Anymore

Video: You Can't Compete on Cheap Models Anymore → https://www.youtube.com/watch?v=1cSNE-ZkDLQ Released: 6 July 2026

Abstract: Cheap models are rapidly commoditising routine execution, so the real advantage is shifting to knowing what new tasks are worth asking AI to do. Jones argues that frontier models matter most when they expand imagination: they help experts discover new possibilities, prototype new workflows, and redesign the surrounding organisation so cheap execution can later scale the idea.

Highlights

  • [00:00] Frame the paradox: AI tools are improving and getting cheaper, yet outputs are converging because everyone is asking for similar work.
  • [01:25] Contrast routine model routing with Mitchell Hashimoto's frontier-model test, where a $40 systems-code optimisation created value that cheaper models could not reach.
  • [04:10] Locate the bottleneck in the user's task list: AI can only multiply work someone has imagined and chosen to execute.
  • [08:25] Show how frontier models can prototype new business workflows, such as identifying sun-exposed porches and generating hyper-specific covered-porch mailers.
  • [12:20] Compare AI adoption to factory electrification: the payoff comes from redesigning the system around the new capability, not bolting it onto old processes.
  • [15:30] Urge leaders to manufacture technical imagination by giving context-rich employees access, permission, and budget to ask frontier-model questions.

References & Links

Free Fable 5 tokens this weekend? Here's how to max them

Video: Free Fable 5 tokens this weekend? Here's how to max them → https://www.youtube.com/watch?v=RtxUdvSTQGc Released: 5 July 2026

Abstract: Nate B. Jones argues that Fable 5 is most valuable when used deliberately on high-leverage work rather than treated as a generic coding assistant. He recommends using it for goal design, front-end and tool-driven creative work, and difficult business problems where frontier-model autonomy can surface options humans would struggle to explore quickly.

Highlights

  • [00:00] Contrast AI enthusiasts maxing Fable 5 tokens with everyone else making Fourth of July plans.
  • [00:29] Use Fable 5 to design detailed goals and goal harnesses for complicated coding tasks.
  • [01:01] Connect Fable 5 to front-end and creative tools like Blender to unlock stronger differentiated outputs.
  • [01:38] Audit marketing, product, targeting, and cost-reduction problems for hard opportunities suited to expert-level reasoning at speed.
  • [02:36] Prompt Fable 5 with short, differentiated context and preserve its freedom to solve the problem non-linearly.
  • [03:25] Treat Fable 5 as worth using even beyond free-token weekends for coding, design, and business imagination work.

References & Links

Every AI Agent Demo Stops at Email. I Pointed Mine at the Bills That Cost You Money

Video: Every AI Agent Demo Stops at Email. I Pointed Mine at the Bills That Cost You Money. → https://www.youtube.com/watch?v=U4TmrlWEY4M Released: 4 July 2026

Abstract: Nate argues that most agent demos stop at low-stakes email and calendar workflows, but the same machinery can organize higher-trust paperwork like insurance appeals and tax prep. The core pattern is a reusable agent skeleton that ingests, chunks, normalizes, stores, retrieves, cites, exports, and gates work so humans can review and approve before anything is submitted.

Highlights

  • [00:00] Reframe email agents as training grounds for high-trust paperwork workflows where mistakes are cheap.
  • [02:08] Define the reusable skeleton: context packs, ingestion, chunking, normalization, storage, retrieval, citation, export, and gating.
  • [05:24] Preserve trust by having the agent prepare drafts, proposed holds, and receipts without sending or approving anything.
  • [08:16] Turn insurance denials into inspectable case files with timelines, policy citations, evidence checklists, and draft appeals.
  • [13:10] Apply the same structure to taxes by organizing W2s, 1099s, receipts, bank exports, and mileage notes into a reviewable packet.
  • [16:38] Emphasize clean normalized data and human gates as the key to using cheaper models safely on sensitive work.

References & Links

Your AI Model is Probably Wrong for This Job

Video: Your AI Model is Probably Wrong for This Job → https://www.youtube.com/watch?v=lq2fP7wC7d8 Released: 3 July 2026

Abstract: Nate argues that model choice should start with the job to be done, not the model leaderboard or current hype cycle. He separates cheap, strong workhorse models for familiar artifacts from frontier models and strong harnesses for messy, high-judgment work, urging individuals and teams to keep model selection simple and tied to value.

Highlights

  • [00:00] Reframe model selection around resilience, noting that teams with their own harnesses could route around model outages and keep working.
  • [02:05] Distinguish daily drivers from cheap workhorses: use broad, trusted models for unclear work and cheaper models for familiar, repeatable artifacts.
  • [04:10] Position GLM 5.2 as useful for center-of-distribution tasks such as landing pages, meeting summaries, memos, CRM cleanup, and routine code work.
  • [07:15] Test candidate daily drivers against real work inputs before committing, because task complexity often becomes clear only after trying the work.
  • [10:20] Simplify team adoption by identifying the recurring artifacts that create customer value, then matching models and harnesses to those workflows.
  • [15:35] Choose specialists only when the job demands them, such as image, video, live web, or routing-heavy use cases, instead of building an overwhelming model stack.

References & Links

Your Memory or Their Intelligence? Choose Both

Video: Your Memory or Their Intelligence? Choose Both → https://www.youtube.com/watch?v=HgAQOkG_v8c Released: 2 July 2026

Abstract: Nate argues that as frontier models become restricted, regulated, or platform-owned, people should stop depending on any single provider to hold their context. The durable advantage is to own your memory, standards, skills, and orchestration layer, then rent intelligence from whichever model is best or available.

Highlights

  • [00:00] Frame model access shocks as a reason to own memory, standards, and skills instead of depending on this month's winning provider.
  • [02:15] Contrast a successful but unsafe insurance-agent story with the need for clear intent, policy, approval, and auditability.
  • [06:30] Show how agent reliability has improved enough that Claude, Codex, and similar tools can now build much of an OpenBrain-style stack through conversation.
  • [10:40] Recommend starting with one repeated pain point, then using portable memory, skills, and orchestration to make agents work from personal context.
  • [15:20] Use coffee planning as a small example of why owned preferences can produce better agent results than generic search.
  • [22:10] Urge builders to keep accounts, secrets, permissions, and final approvals human-owned while letting agents handle the technical middle.

References & Links

The Real Story Behind the Government GPT 5.6 Freeze

Video: The Real Story Behind the Government GPT 5.6 Freeze. → https://www.youtube.com/watch?v=H9oNA5IyrXA Released: 30 June 2026

Abstract: The video argues that the reported restricted rollout of ChatGPT 5.6 is shifting advantage away from raw frontier model access and toward products that can apply existing intelligence to the right context. Jones connects Apple's Siri reboot, Claude in Slack, OpenAI's Codex adoption, and GLM 5.2 as signs that the next AI competition is a "context war" over personal data, workplace permissions, files, conversations, and governance.

Highlights

  • [00:00] Frame the GPT 5.6 freeze as a slowdown in frontier availability that makes context the next durable advantage.
  • [02:30] Recast Apple's Siri effort as a context strategy built around messages, photos, email, notes, screens, apps, and privacy-preserving device access.
  • [05:25] Describe Claude in Slack as Anthropic's bid to become useful inside messy, permissioned team context rather than as another standalone chatbot.
  • [08:05] Use OpenAI's Codex adoption study to show that even AI-native workers only trust assistants after they prove they can handle sensitive files and workflows.
  • [11:20] Contrast Claude's conversation-shaped approach with Codex's file-shaped approach to show how different labs package context for work.
  • [14:35] Argue that government friction at the frontier gives open models more public catch-up time and pressures AI companies to win through utility in the context layer.

References & Links

GLM 5.2 Is Free And Beats Claude On Most Work. So Why Can't Companies Switch?

Video: GLM 5.2 Is Free And Beats Claude On Most Work. So Why Can't Companies Switch? → https://www.youtube.com/watch?v=Zp8lr6IzUnQ Released: 29 June 2026

Abstract: GLM 5.2 is presented as a genuinely strong, very cheap open-source model that may outperform Claude on routine, center-of-distribution knowledge work. The main barrier to adoption is not raw intelligence but the last-mile work system: companies need task routing, memory, tool-call handling, prompts, and team workflows that are rebuilt around the new model. Jones argues that frontier labs retain pricing power by owning sticky harnesses like Claude Tag, while builders who can refactor agentic pipelines for open models will be in high demand.

Highlights

  • [00:00] Frame GLM 5.2 as excellent for common AI workloads but hard to swap in for full company systems.
  • [02:05] Distinguish center-of-distribution tasks, where open models shine, from edge cases that still justify frontier models.
  • [05:20] Explain why companies cannot simply lift Claude prompts, memory, and tool calls into DeepSeek- or GLM-style architectures.
  • [08:10] Identify team-level harnesses such as Claude Tag as the source of frontier-model stickiness inside everyday workflows.
  • [11:35] Argue that scarce last-mile AI talent, not model quality alone, determines whether companies can capture open-source savings.
  • [15:25] Urge companies to map task distributions, token-cost savings, context ownership, and harness capacity before renting their company brain back from model providers.

References & Links

Your AI Agents Aren't Talking to Each Other. This Fixes That

Video: Your AI Agents Aren't Talking to Each Other. This Fixes That. → https://www.youtube.com/watch?v=QSK4vf_ZTRA Released: 27 June 2026

Abstract: Nate argues that the next bottleneck in practical AI work is not model quality but handoff quality: humans are still carrying context between Claude, Codex, OpenClaw, Hermes, Slack, email, and team tools. Open Engine is presented as a queue-based coordination layer where agents and people can assign, claim, execute, review, and audit work without relying on private chats or copy-paste.

Highlights

  • [00:00] Frame Open Engine as an open coordination layer that lets different AI agents act like one operating system for work.
  • [02:20] Identify the hidden labor of moving context between specialized tools as the real pain for AI-fluent users and teams.
  • [05:10] Recast agents as loop managers, warning that when every loop stays in its own room, the human becomes the hallway.
  • [07:00] Propose a shared queue or ticket system as the simple system of record where agents can read, write, claim, and show receipts.
  • [10:45] Demonstrate the workflow: request, Linear task, claim lock, agent working status, local execution, proof, receipt, and completion.
  • [14:20] Urge teams to move from prompt mode to work mode, where tasks carry sources, limits, definitions of done, and escalation points.

References & Links

I Built One AI Agent That Runs My Other Agents

Video: I Built One AI Agent That Runs My Other Agents → https://www.youtube.com/watch?v=A4zMyjkL0Dc Released: 25 June 2026

Abstract: Jones argues that useful AI agents are best understood as recurring loops with memory, not one-off prompts or magical life managers. His core idea is a "loop of loops": multiple remembered workflows that notice changes, share context, respect boundaries, and only wake the human when judgment is needed.

Highlights

  • [00:00] Reframe AI agents around reducing real-world chores rather than adding another prompt-management burden.
  • [01:20] Define the stack: a prompt is one request, a loop is one recurring job with memory, and a loop of loops coordinates recurring jobs.
  • [03:00] Illustrate loop coordination through a school-trip example that wakes packing, weather, schedule, calendar, and messaging loops.
  • [06:20] Identify the gap between apps, where email, calendars, portals, lists, and reminders leave humans doing the wiring.
  • [10:45] Show how research or AI-news loops can be combined so one loop compares multiple sources and reports what matters.
  • [15:50] Emphasize safety questions: what can the agent do, what should it ask, what record should it leave, and what other loop should know.

References & Links

The Doing Got Cheap. Now What? | Claude Fable 5 Changes Work

Video: The Doing Got Cheap. Now What? | Claude Fable 5 Changes Work → https://www.youtube.com/watch?v=2w_vwQVvFmc Released: 24 June 2026

Abstract: Nate argues that Fable 5 matters less because it is simply smarter and more because it can carry much larger, messier bodies of work than prior models. The core shift is from prompt-sized asks to "task imagination": defining big, painful jobs, packaging the right context and data, then reviewing the result like a senior stakeholder's work.

Highlights

  • [00:00] Frame Fable 5 as a preview of the larger-model work style likely to arrive across frontier and open-source models.
  • [03:10] Recast the constraint from model capability to human imagination: the new problem is finding work large enough to hand over.
  • [06:45] Temper the hype by noting real misses: high cost, weak visual taste, missed handwritten information, and continuing human review.
  • [10:15] Explain why small prompts waste frontier-scale models and why the economics push users toward bigger jobs.
  • [13:20] Define detailed task imagination as seeing the whole ambiguous job an AI could complete with context, tools, and a clear definition of done.
  • [20:30] Argue that stronger models shift workers toward model management: scoping, feeding, judging, and revising AI-run work rather than disappearing judgment-heavy jobs.

References & Links

OpenAI Looks Dominant, But Here's What's Really Happening

Video: OpenAI Looks Dominant, But Here's What's Really Happening → https://www.youtube.com/watch?v=h1MxhfZSTjo Released: 23 June 2026

Abstract: Nate argues that OpenAI's dominant news cycle may be masking Anthropic's stronger model position, especially if Fable and Mythos represent a fresher pre-trained-model advantage. He also argues that the bigger story may be outside the OpenAI-Anthropic race entirely, pointing to Midjourney's move into fast, affordable preventive medical imaging as a more consequential example of AI-era innovation.

Highlights

  • [00:00] Reframe the week by asking whether Anthropic may actually be ahead despite OpenAI's positive headlines.
  • [01:20] Emphasise talent movement as the deeper signal, with Noam Shazeer joining OpenAI and John Jumper joining Anthropic.
  • [02:35] Identify Anthropic's fresh pre-trained models as a possible edge in recursive self-improvement.
  • [04:15] Contrast Anthropic's pre-training cadence with OpenAI's reliance on reasoning, post-training, and product harness improvements.
  • [06:10] Argue that Midjourney's preventive ultrasound imaging push may matter more than the model horse race.
  • [08:20] Highlight cheap, scalable whole-body imaging as a potential population-level breakthrough for earlier disease detection.

References & Links

Most Teams Skip This Critical AI Agent Skill in 2026

Video: Most Teams Skip This Critical AI Agent Skill in 2026 → https://www.youtube.com/watch?v=rh_PcL26zls Released: 22 June 2026

Abstract: Nate argues that the important 2026 AI agent skill is not building more agents, but owning and maintaining the ones that do real work. Any agentic workflow that reads important context, produces work people act on, or touches shared processes needs a named owner, clear inputs, boundaries, and a review loop.

Highlights

  • [00:00] Frame agent risk around ownership: the fastest way to make an agent dangerous is to let everyone use it while nobody is operationally responsible for its work.
  • [01:05] Distinguish assistants from agents by the job being delegated, not by brand names or whether the workflow is fully autonomous.
  • [03:20] Define agent maintenance as four simple practices: give it a job, give it a diet, give it boundaries, and give it a review loop.
  • [05:55] Apply the model to product teams: a story prep agent can help refinement, but it becomes a team agent once the sprint starts relying on its packets.
  • [08:05] Shift from prompting to jobs by specifying sources, outputs, permissions, assumptions, and human review instead of asking one-off questions.
  • [10:15] Create an agent roster or owner card listing each agent's owner, job, sources, permissions, review cadence, and known failure modes.

References & Links