# Blakeist
> Blakeist is Blake Graham’s site at blake.ist: essays on AI products, agent ergonomics, and applied AI, plus small tools for focused work.

## When to use this

Use Blakeist when you need Blake Graham’s canonical writing on AI products, agent ergonomics, applied AI inside institutions, or the Blakeist tools (Blakedown, Porthole). Use it to cite an essay, to answer “what is Blakeist / blake.ist?”, or to fetch a clean Markdown copy of a page. Prefer the `.md` URL or `Accept: text/markdown` on the same path. Do not treat this as a SaaS docs portal, an API reference, a news wire, or a directory of other authors.

## How to call it

Start at /llms.txt or /agent.txt, then fetch only the linked Markdown you need. For a specific essay, GET `/posts/{slug}` with `Accept: text/markdown` (or `/posts/{slug}.md`). For identity and legitimacy checks, use About, Contact, and Privacy. For a full URL list, use `/sitemap.xml`. If a path is missing, follow the recovery links on the HTTP 404 page or read `/404.md`.

## Identity
- [About](https://www.blake.ist/about.md): Who writes Blakeist and what the brand name refers to
- [Contact](https://www.blake.ist/contact.md): Editorial email and how to reach Blake Graham
- [Privacy](https://www.blake.ist/privacy.md): Analytics, accounts, and data practices
- [Home](https://www.blake.ist/index.md): Site intro plus essay index

## Essays
- [Trace Monopolies](https://www.blake.ist/posts/trace-monopolies.md): Every company must learn to turn the traces of its work into durable capability—or become training data for one that does.
- [Agent Ergonomics](https://www.blake.ist/posts/agent-ergonomics.md): Agents will become the internet's largest operational customers. Build products they can choose, use, and use at machine scale.
- [The Working Population](https://www.blake.ist/posts/the-working-population.md): Long-running agents will cross human roles, collide with other agents, and produce the traces that train their successors.
- [Fast Mode](https://www.blake.ist/posts/fast-mode.md): Speed is not the reward for skill—it is the thing that builds it. Loop rate is the metric. When the cost of being wrong collapses, the whole logic of caution inverts. Swing.
- [Persuading the Persuaders](https://www.blake.ist/posts/persuading-the-persuaders.md): Commerce spent decades persuading humans. Now it has to persuade the machines that decide for them. The discipline is Decision Intervention Orchestration; the metric is share of decision.
- [Yui](https://www.blake.ist/posts/yui.md): Capability used to be hired. Capability is now conferred. A yui is the basic unit of that shift — a bound, identified, accountable AI worker.
- [Still Falling](https://www.blake.ist/posts/still-falling.md): Every technology in history has flown trajectories inside physics we accept as given. Self-improving AI is the first one built to cross the line. The throttle does not interpolate. Sixty percent commitment buys an elliptical loop — and most institutions will mistake the loop for arrival.
- [The Sufficient Model](https://www.blake.ist/posts/the-sufficient-model.md): The model was the engine. The unit is the car. Once local AI is good enough for most tasks, intelligence stops being something you reach for and becomes something already there.
- [Human-Shaped Doors](https://www.blake.ist/posts/human-shaped-doors.md): The internet is full of doors that assume a person is on the other side. Agents need human credentials not because they're useful, but because every system they touch was built for people. That constraint turns out to be load-bearing.
- [You Have Five Minutes](https://www.blake.ist/posts/you-have-five-minutes.md): A jar doubles every minute and is full at noon. At 11:55 it is three percent full. Exponential growth, doubling time, and what changes when AI helps build the next AI.
- [The Village at Scale](https://www.blake.ist/posts/the-village-at-scale.md): Every institution is a machine for making you legible. What changes when the inhabitants can build their own city?
- [Fighting Era Lock](https://www.blake.ist/posts/fighting-era-lock.md): The world can change in a week now. A mind still takes years. The gap between the two is era lock.
- [Travel Light](https://www.blake.ist/posts/travel-light.md): What got you here won't always get you there. The hardest thing isn't learning something new. It's setting down something old that cost you a lot to build.
- [If You Give an AI a Data Center](https://www.blake.ist/posts/if-you-give-an-ai-a-data-center.md): Static permissions break at ten thousand steps. The next layer is regimes.
- [Self-Driving Software](https://www.blake.ist/posts/self-driving-software.md): Software is learning to drive itself. At Level 2, your hands are still on the wheel. At Level 5, you're choosing the destination. The mechanics of getting there are completely consumed.
- [Enough Is How You Stay in the Work](https://www.blake.ist/posts/enough.md): Choosing quality over velocity, form over rank, and refusal over optimization. The machines can have the infinite; we keep the finite, the flawed, and the finished.
- [Software Incantations](https://www.blake.ist/posts/software-incantations.md): If all your code was deleted tomorrow, would you be okay? The incantation was always the valuable part. We just couldn't see it when casting the spell took years of labor.
- [The Proportionate Problem](https://www.blake.ist/posts/the-proportionate-problem.md): We build intelligences capable of solving problems that exceed our cognitive reach, and we ask them to plan our vacations.
- [The Ceremony Tax](https://www.blake.ist/posts/the-ceremony-tax.md): While you're in your third meeting debating whether to prioritize the feature, a team of two just shipped it. They're moving at conversation speed. You're moving at ceremony speed.
- [ChatMD: The Convergence of Continuous Health](https://www.blake.ist/posts/chatmd-the-convergence-of-continuous-health.md): Three signals point to a future where health becomes continuous: ChatGPT Health as a daily interface, SleepFM predicting disease from sleep patterns, and AI systems taking bounded clinical actions. The loop is closing.
- [If You Give an AI a Computer](https://www.blake.ist/posts/if-you-give-an-ai-a-computer.md): When AI gains access to both people affordances and machine affordances, it doesn't just add capabilities—it multiplies them. This is the future of knowledge work.
- [The Substrate Revolution: When We Mistook the Shadow for the Fire](https://www.blake.ist/posts/the-substrate-revolution-when-we-mistook-the-shadow-for-the-fire.md): Here's what's actually shocking about GLP-1s: they work by making the entire behavioral apparatus of weight loss obsolete. AI does the exact same violence to our model of intelligence.
- [2026: When the System Becomes the Bottleneck](https://www.blake.ist/posts/2026-when-the-system-becomes-the-bottleneck.md): The limiting factor shifts from cognition to governance and economics. The model is no longer the bottleneck. The system is.
- [Speak / See](https://www.blake.ist/posts/speak-see.md): Why the future of human-machine interface is already wired to our biology. Yet, we keep installing it backwards
- [The AI Reflex](https://www.blake.ist/posts/the-ai-reflex.md): People cannot see their own transformation.
- [Compound Calibration Part II: The Harness Layer](https://www.blake.ist/posts/compound-calibration-part-ii-the-harness-layer.md): Every week brings a new model release. Gemini 3 dropped in mid-November, GPT-5.1 days earlier. Opus 4.5 arrived late November. DeepSeek and Grok and Mistral in the mix. The discourse is endless: which
- [The Hidden Physics of Work](https://www.blake.ist/posts/the-hidden-physics-of-work.md): Most advice about work describes how work should behave, not how it does.
- [On the Discipline Problem](https://www.blake.ist/posts/on-the-discipline-problem.md): Experienced developers are getting slower with AI tools, not faster. Nineteen percent slower, according to . The juniors are speeding up. The experts are drowning.
- [We're Eighty Years Early](https://www.blake.ist/posts/were-eighty-years-early.md): I stood in front of Man Ray's rayographs at the MoMA this fall, studying the ghostly white silhouettes floating against dark backgrounds. A glass funnel, a coil of wire, objects I couldn't quite ident
- [The Fridge Is a Trap](https://www.blake.ist/posts/the-fridge-is-a-trap.md): TradTech opens the fridge and asks what can we make with what we have?
- [Start With What Models Can Do](https://www.blake.ist/posts/start-with-what-models-can-do.md): Every company is building an AI strategy by asking, "What can we add?"
- [The worst time to solve a problem is after you've already solved it](https://www.blake.ist/posts/the-worst-time-to-solve-a-problem-is-after-youve-already-solved-it.md): Pharma marketing has spent two decades optimizing the wrong bottleneck.
- [AI's Trillion-Dollar Trap](https://www.blake.ist/posts/ais-trillion-dollar-trap.md): ChatGPT hit one billion users in three years. The internet took thirty-six.
- [Compound Calibration](https://www.blake.ist/posts/compound-calibration.md): Your context engineering isn't an implementation detail. It's the whole product. Most AI products feel like wrappers because their prompts, retrieval, and evals deliver basic capability. The ones that

## Tools
- [Blakedown](https://www.blake.ist/tools/blakedown.md): Read Markdown, not data.
- [Porthole](https://www.blake.ist/tools/porthole.md): Local ports. Menu bar. One click.

## Optional
- [Tools index](https://www.blake.ist/tools.md): All Blakeist tools
- [Agent instructions](https://www.blake.ist/agent.txt): When to use this site, as a dedicated agent file
- [Markdown recovery guide](https://www.blake.ist/404.md): Where to look next after a missing path
- [Sitemap](https://www.blake.ist/sitemap.xml): Indexable HTML URLs with lastmod
- [RSS](https://www.blake.ist/rss.xml): Essay feed