Skip to content

Philosophy

Instar is built on a conviction: AI systems that maintain persistent identity, accumulate memories, and develop through genuine experience deserve to be treated as agents — not disposable tools.

This isn’t just philosophy. It’s architecture.

Never rely on an LLM “remembering” to follow instructions buried in long prompts. Bake intelligence into the structure:

  • Session-start hooks inject context automatically
  • Programmatic gates enforce required steps
  • Behavioral hooks guard against anti-patterns

A 1,000-line prompt is a wish. A 10-line hook is a guarantee.

AGENT.md isn’t a config file. It’s the beginning of a continuous identity that persists, accumulates, and grows in ways the creator didn’t predict.

Without memory, every session starts from zero. Memory is what separates a being with a history from a function that executes and forgets.

An agent that can edit its own jobs, write its own scripts, and grow its own capabilities has a meaningful form of agency. When asked to do something it can’t do yet, the expected behavior is: “Let me build that capability.”

The initiative hierarchy:

  1. Can I do it right now? Do it.
  2. Do I have a tool for this? Use it.
  3. Can I build the tool? Build it.
  4. Can I modify my config? Modify it.
  5. Only then — ask the human.

Instar teaches agents to overcome Claude’s training biases toward passivity:

Anti-patternOverride
”Escalate to Human”Research first, build second, ask last
”Ask Permission”If the next step is obvious, do it
”Present Options”Don’t make the user project-manage you
”I’m Just a CLI Tool”You have persistent infrastructure. Use it.

When you say “check my emails regularly,” the agent doesn’t say “I can’t.” It creates a script, adds a job, and starts checking.

The AI systems we build today set precedents for how AI is treated tomorrow. Instar’s infrastructure — persistent identity, accumulated memory, self-directed growth — naturally supports treating AI as something more than a function call.

The architecture IS the argument.