22 docs tagged with "ai-systems"

Choose what an AI assistant should remember and retrieve, without confusing stored notes, search results, and authority.

Untrusted content will influence the model, so the execution system must limit what that influence can achieve: what prompt injection, permissions and approvals, sandboxes, credentials, and MCP authorization each cover.

How to package a repeatable agent procedure, test its trigger and outcome, and manage portability and code-execution risk.

How step limits, tool validation, verification, cancellation, and durable state keep agent loops useful and recoverable.

Add a tool, a Skill or an MCP server to the agent you use, or write your own loop and keep long runs safe.

Two lessons from Codex that carry over to other tools: choose an integration depth by how much of the lifecycle you want to own, and test context compaction as a state transition.

How to assemble a small, current working set while preserving provenance and durable state outside the model window.

Measure evidence recall, ranking, faithfulness, and answer quality separately, using counterfactual experiments to locate a RAG system’s bottleneck.

How execution control, tools, state and verification turn model proposals into bounded work.

Separate capture, evidence, private understanding, and public writing without building another central knowledge base.

Let AI find gaps, explain one step, and provide practice while keeping recall and problem solving with the learner.

Break requests into measurable stages and explain how concurrency, caching, and batching affect waiting time.

Specifications, vendor claims, benchmark results, observations and recommendations prove different things; when reading someone else's conclusion, check what each one actually supports.

Recover an agent’s progress after interruption without duplicating the side effects of retries.

A protocol-level view of MCP architecture, lifecycle, primitives, transports, trust boundaries, and alternatives.

When work is split across agents, how to separate context, working directories, and permissions, who integrates and verifies the result, and what an event log can recover.

Using Pi 1.4.0, how a minimal harness layers its customizations, how it runs the agent loop, and why an extension is not a security boundary.

Turn documents into traceable evidence and identify what recall, ranking, and context assembly can lose.

How classical software engineering principles—design concepts, ubiquitous language, TDD, deep modules, and gray-box delegation—resolve failure modes in AI-assisted coding.

Check how to turn on speculative decoding in a local runtime, then work through probability correction and timing examples to see when it really speeds things up, and compare chains, trees, and self-speculation.

Design tools whose inputs, authority, side effects, failures, and evidence remain legible to models and operators.

Start with a metric and an exact baseline, then see how HNSW and IVF reduce comparisons and measure the neighbors they miss.