Notes on AI products, agents, memory, systems, and the craft of building.https://2ander.com/https://2ander.com/writing/agents-should-earn-autonomy/https://2ander.com/writing/agents-should-earn-autonomy/Autonomy is not a switch from low to high. It is a set of decision rights that must be granted, evaluated, constrained, and revoked independently.Fri, 14 Aug 2026 00:00:00 GMThttps://2ander.com/writing/agents-md-and-lightweight-specs/https://2ander.com/writing/agents-md-and-lightweight-specs/Long-running collaboration with coding agents needs durable rules and explicit decisions, but not a specification for every change.Thu, 13 Aug 2026 00:00:00 GMThttps://2ander.com/writing/coding-agents-as-execution-systems/https://2ander.com/writing/coding-agents-as-execution-systems/A technical walk through how a coding agent explores, plans, edits, verifies, and recovers while fixing a bug across UI, API, and persistence.Wed, 12 Aug 2026 00:00:00 GMThttps://2ander.com/writing/the-hard-part-is-state-authority/https://2ander.com/writing/the-hard-part-is-state-authority/Reliable agent products begin by deciding which system owns each fact—not by making the reasoning loop more elaborate.Sat, 08 Aug 2026 00:00:00 GMThttps://2ander.com/writing/enterprise-ai-delivers-artifacts-not-answers/https://2ander.com/writing/enterprise-ai-delivers-artifacts-not-answers/The value of enterprise AI is not a more fluent conversation. It is turning ambiguous work into outcomes that can be reviewed, executed, and trusted.Sat, 25 Jul 2026 00:00:00 GMThttps://2ander.com/writing/context-memory-and-attention-budget/https://2ander.com/writing/context-memory-and-attention-budget/Reliable agent context depends on separating archives, durable memory, and the current working set—not loading everything the system knows.Sun, 12 Jul 2026 00:00:00 GMThttps://2ander.com/writing/from-model-calls-to-agent-systems/https://2ander.com/writing/from-model-calls-to-agent-systems/How a model call grows into an agent loop, and why reliable execution requires runtime, context, state, permissions, and verification.Fri, 10 Jul 2026 00:00:00 GMThttps://2ander.com/writing/memory-is-a-product-surface/https://2ander.com/writing/memory-is-a-product-surface/Useful memory is not a larger transcript. It is a product’s ability to preserve continuity without accumulating confusion.Thu, 02 Jul 2026 00:00:00 GMThttps://2ander.com/writing/billing-boundaries-in-generative-ai/https://2ander.com/writing/billing-boundaries-in-generative-ai/Reliable billing depends on defining exactly when work is accepted, completed, delivered, settled, or reversed—not only on maintaining a price table.Wed, 24 Jun 2026 00:00:00 GMThttps://2ander.com/writing/model-attempts-behind-a-successful-request/https://2ander.com/writing/model-attempts-behind-a-successful-request/A request can succeed for the user while the infrastructure behind it is already failing. The missing unit of observation is the individual model attempt.Fri, 12 Jun 2026 00:00:00 GMThttps://2ander.com/writing/a-system-should-not-wait-to-be-asked/https://2ander.com/writing/a-system-should-not-wait-to-be-asked/A proactive personal system must know what deserves an interruption, when to stay silent, and how outcomes should change its next judgment.Sat, 18 Apr 2026 00:00:00 GMThttps://2ander.com/writing/media-generation-is-not-an-api-call/https://2ander.com/writing/media-generation-is-not-an-api-call/Long-running media generation is a distributed job that needs durable state, recovery, settlement, and delivery—not a single API call.Sat, 28 Mar 2026 00:00:00 GMThttps://2ander.com/writing/tool-calling-is-not-an-agent-runtime/https://2ander.com/writing/tool-calling-is-not-an-agent-runtime/Native tool calling made Oracle's decisions cleaner, but reliability still depended on state, inheritance, verification, and recovery.Fri, 30 Jan 2026 00:00:00 GMThttps://2ander.com/writing/why-squad-did-not-work/https://2ander.com/writing/why-squad-did-not-work/Squad could run and replan for months. It still could not evaluate whether today's action would remain correct over a long horizon.Fri, 14 Nov 2025 00:00:00 GMThttps://2ander.com/writing/agents-should-outlive-the-request/https://2ander.com/writing/agents-should-outlive-the-request/Model inference happens inside a call, but real work unfolds over time. A reliable agent must pause, recover, remain visible, and turn completion into delivery.Thu, 28 Aug 2025 00:00:00 GMThttps://2ander.com/writing/why-we-built-oracle/https://2ander.com/writing/why-we-built-oracle/When a user gives AI a goal rather than a question, the fundamental unit of the product must shift from the answer to the task.Fri, 23 May 2025 00:00:00 GMThttps://2ander.com/writing/how-to-really-give-a-webpage-to-ai/https://2ander.com/writing/how-to-really-give-a-webpage-to-ai/A URL is not content, and copy-paste is not the answer. A web page has to travel through an easily overlooked path before AI can use it as material.Fri, 27 Dec 2024 00:00:00 GMT