Research

Notes from the frontier

Explorations at the edge of cognitive infrastructure, autonomous agents, and intelligent systems.

3 posts · filter by topic

Featured
AgentsHarnessesArchitecture

The Agent Harness: State of the Art (2024–2026)

An LLM by itself is a stateless function: it receives tokens and emits tokens. Everything that turns that function into something that does work over time — the loop, the tools, the memory, the sandbox, the guardrails — is the harness. This post surveys the field, with evidence.

Aug 26, 202615 min readAbel Santillan Rodriguez
Read more
AgentsHarnesses

The New Agent Harnesses, Compared (2024–2026)

Claude Code, Codex, Pi, OpenCode, DeepSeek dsh, Qwen Code, Aider, Cline/Roo, Gemini CLI, Goose, Jules, mini-SWE-agent and more — each makes an explicit design bet about how a coding agent should be built. This is the comparison, with the numbers.

Aug 25, 202615 min
BenchmarksAgents

Beyond the 5-Minute Task: Why Current Agent Benchmarks Are Failing Us

Most existing benchmarks measure performance on tasks completable in minutes. Real-world deployment demands agents that operate reliably over hours, days, or weeks. This post surveys the gaps and proposes a framework for Long-Time-Task benchmarks.

Apr 1, 20259 min