Field notes
Planning, tool use, evaluation, and the boundaries of autonomous software. Browse the connected projects, notes, and references below.
topics / JGARDEN
AI & agents
Planning, tool use, evaluation, and the boundaries of autonomous software.
Connected to this topic
6 entries
✳sapling
Agents need boundaries, not just better prompts
A working note on giving coding agents explicit authority, budgets, and a definition of done.
agentsreliabilitytooling
✳evergreen
Designing systems you can replay
Deterministic inputs make failures easier to investigate, but external effects still need boundaries.
eventsreliability
▤blog
Evidence before autonomy
A useful engineering agent should make its work easier to inspect.
agentsevaluation
⌘research
AI Development Engine
A foundation for planning, executing, and reviewing bounded software-engineering agent tasks.
PythonOllamaPostgreSQL
⌘research
Agent Operating System
Exploring shared memory, planning, execution, and coordination for a network of agents.
PythonEvents
⌘research
Repository Intelligence System
Turn source code, dependencies, and change history into evidence for engineering decisions.
PythonGitTree-sitter
No entries match these filters. Try a different topic or reset the filters.
Continue exploring
▤blog
Evidence before autonomy
A useful engineering agent should make its work easier to inspect.
agentsevaluation
✳sapling
Agents need boundaries, not just better prompts
A working note on giving coding agents explicit authority, budgets, and a definition of done.
agentsreliabilitytooling
✳evergreen
Designing systems you can replay
Deterministic inputs make failures easier to investigate, but external effects still need boundaries.
eventsreliability