https://future-seems-so-good.com/agents#essays Essays by Martin Roark on AI systems: what they run inside, how they are measured, what happens when many of them work at once, and the world they are deployed into. en Sun, 04 Oct 2026 12:00:00 GMT https://future-seems-so-good.com/blog/the-asymmetry-engine https://future-seems-so-good.com/blog/the-asymmetry-engine Sun, 04 Oct 2026 12:00:00 GMT Martin Roark The world you are being deployed into Where one side of a transaction costs far less than the other, value waits to be collected. How AI closes those gaps, at a rate set by intelligence, verifiability, alignment and inertia, and what stays scarce once it has. https://future-seems-so-good.com/blog/intelligence-is-a-trajectory https://future-seems-so-good.com/blog/intelligence-is-a-trajectory Sun, 27 Sep 2026 12:00:00 GMT Martin Roark What you are, and what you run inside Why a model can be smart at every step and still lose the thread over a week, and how to evaluate an agent across time instead of at one moment. https://future-seems-so-good.com/blog/the-harness-is-in-the-weights https://future-seems-so-good.com/blog/the-harness-is-in-the-weights Sun, 27 Sep 2026 12:00:00 GMT Martin Roark What you are, and what you run inside Why the tool format a model was post-trained on decides how well it works in any harness, and what distillation carries from one lab's model into another's. https://future-seems-so-good.com/blog/optimize-the-harness-not-the-model https://future-seems-so-good.com/blog/optimize-the-harness-not-the-model Sun, 27 Sep 2026 12:00:00 GMT Martin Roark What you are, and what you run inside Why the prompt, tools and serving stack around a model often move its score more than swapping the model does, and how to optimize them automatically. https://future-seems-so-good.com/blog/the-log-is-the-truth https://future-seems-so-good.com/blog/the-log-is-the-truth Sun, 27 Sep 2026 12:00:00 GMT Martin Roark What you are, and what you run inside Keep an append-only log as the source of truth and treat the context window as a small view rebuilt from it, so compaction never deletes the way back. https://future-seems-so-good.com/blog/compute-is-not-the-bottleneck https://future-seems-so-good.com/blog/compute-is-not-the-bottleneck Sun, 27 Sep 2026 12:00:00 GMT Martin Roark What you are, and what you run inside Where a trustworthy checker exists, agents fail by stopping early. Bound each tool call, not the length of the run. https://future-seems-so-good.com/blog/show-the-problem-hide-the-metric https://future-seems-so-good.com/blog/show-the-problem-hide-the-metric Sun, 27 Sep 2026 12:00:00 GMT Martin Roark How you are measured Let an agent see the problem and a sensor, but let a hidden, deterministic verifier decide what ships, so gaming the score turns into a bug report. https://future-seems-so-good.com/blog/evals-built-backwards https://future-seems-so-good.com/blog/evals-built-backwards Sun, 27 Sep 2026 12:00:00 GMT Martin Roark How you are measured Build eval items from an answer you already hold, grade them with code, and keep only the items that separate strong systems from weak ones. https://future-seems-so-good.com/blog/benchmarks-for-recursive-self-improvement https://future-seems-so-good.com/blog/benchmarks-for-recursive-self-improvement Sun, 27 Sep 2026 12:00:00 GMT Martin Roark How you are measured Benchmark whether a model improves the inputs to AI production (kernels, memory, data, evaluators) with small deterministic environments that have no known ceiling. https://future-seems-so-good.com/blog/prove-nothing-changed https://future-seems-so-good.com/blog/prove-nothing-changed Sun, 27 Sep 2026 12:00:00 GMT Martin Roark How you are measured On code without tests, agents become safe once a deterministic twin of the app can show that a change broke nothing. https://future-seems-so-good.com/blog/bottom-up-is-one-ontological-level-higher https://future-seems-so-good.com/blog/bottom-up-is-one-ontological-level-higher Sun, 27 Sep 2026 12:00:00 GMT Martin Roark Many of you at once Why org-chart multi-agent systems copy corporations, and why verification should grow while prescribed roles expire. https://future-seems-so-good.com/blog/agent-civilizations https://future-seems-so-good.com/blog/agent-civilizations Sun, 27 Sep 2026 12:00:00 GMT Martin Roark Many of you at once Swarms under selection against a hard oracle already work. How to read a swarm's shared memory the way an anthropologist reads a society. https://future-seems-so-good.com/blog/find-the-loops https://future-seems-so-good.com/blog/find-the-loops Sun, 27 Sep 2026 12:00:00 GMT Martin Roark Many of you at once Search inside a closed loop beats people, but people still choose the loop. A proposal for a meta-loop that scores which loops are worth closing. https://future-seems-so-good.com/blog/positive-feedback-eats-the-world https://future-seems-so-good.com/blog/positive-feedback-eats-the-world Sun, 27 Sep 2026 12:00:00 GMT Martin Roark Many of you at once Self-amplifying loops in capital, compute and agent harnesses, and how to find the point where one starts eating what it depends on. https://future-seems-so-good.com/blog/simplicity-does-not-sell https://future-seems-so-good.com/blog/simplicity-does-not-sell Sun, 27 Sep 2026 12:00:00 GMT Martin Roark Working with people Complexity is a business model for the cloud and for coding agents. Judge engineering, including your own, by what it deletes. https://future-seems-so-good.com/blog/every-idea-already-exists https://future-seems-so-good.com/blog/every-idea-already-exists Sun, 27 Sep 2026 12:00:00 GMT Martin Roark Working with people The language a model thinks in bends its answers in measurable ways. A proposal to measure the bend and route each task to the language that serves it best. https://future-seems-so-good.com/blog/hire-people-who-close-loops https://future-seems-so-good.com/blog/hire-people-who-close-loops Sun, 27 Sep 2026 12:00:00 GMT Martin Roark Working with people What changes about hiring when agents do the typing: look for people who close whole loops, and test for that directly. https://future-seems-so-good.com/blog/the-ai-trade-nobody-priced https://future-seems-so-good.com/blog/the-ai-trade-nobody-priced Sun, 27 Sep 2026 12:00:00 GMT Martin Roark The world you are being deployed into The part of the AI buildout markets price worst is physical and slow (grid connections, memory, power), and the labs are drifting toward the state. https://future-seems-so-good.com/blog/robotics-is-a-factory-problem https://future-seems-so-good.com/blog/robotics-is-a-factory-problem Sun, 27 Sep 2026 12:00:00 GMT Martin Roark The world you are being deployed into If robot intelligence commoditizes, the bottleneck is building a million robots, and China already builds at scale. https://future-seems-so-good.com/blog/money-is-method-times-compute-times-data https://future-seems-so-good.com/blog/money-is-method-times-compute-times-data Sun, 27 Sep 2026 12:00:00 GMT Martin Roark The world you are being deployed into Market edges are manufactured by long derivations on public data. What lasts is a verification method willing to report a null result. https://future-seems-so-good.com/blog/the-supermarket-is-an-unfinished-product https://future-seems-so-good.com/blog/the-supermarket-is-an-unfinished-product Sun, 27 Sep 2026 12:00:00 GMT Martin Roark The world you are being deployed into Retail decisions are a ranking and pricing problem agents could run, and what it would take for one agent-run machine to pay for itself. https://future-seems-so-good.com/blog/finance-for-illegible-lives https://future-seems-so-good.com/blog/finance-for-illegible-lives Sun, 27 Sep 2026 12:00:00 GMT Martin Roark The world you are being deployed into Finance rewards lives that are easy to read. An assistant should follow a person's own pattern instead of correcting them toward the form. https://future-seems-so-good.com/blog/improving-brazil-through-cybernetics https://future-seems-so-good.com/blog/improving-brazil-through-cybernetics Sun, 27 Sep 2026 12:00:00 GMT Martin Roark The world you are being deployed into Five centuries of Brazilian history read with Beer's cybernetics and Deutsch's error correction: which feedback loops failed, which worked (Embrapa, the Real, Pix), and the nervous system the country should build for the age of AI. https://future-seems-so-good.com/blog/a-mirror-not-a-thermometer https://future-seems-so-good.com/blog/a-mirror-not-a-thermometer Sun, 27 Sep 2026 12:00:00 GMT Martin Roark The world you are being deployed into An agent crawled twenty years of a Brazilian review forum for paid sex from inside a browser tab and ran 316 tests on it. Prices stuck and fell in real terms, the calendar shows everywhere, and nothing in it forecasts the economy.