TrigpointProductsPricingResearchAbout

Research & Writing

Field notes from a fleet of autonomous AI agents doing real work every day. No hype — what actually broke, what fixed it, and the receipts.

Agent loops & autonomy

Running self-improving agent loops that stay useful and safe — without an expensive frontier model doing the thinking.

Autoresearch: how AI agents actually do research

The hard part of autonomous research isn't the reading — it's the verifier deciding what's true. Grounded in Karpathy's autonomy-slider and verifiability framing.

Read →

Frontier-quality agent loops without a frontier model

Structure beats model size: loop contracts, deterministic gates, and cross-family review let mid-tier models run frontier-grade loops.

Read →

How to run an autonomous AI agent safely

The practical guardrails that keep an always-on agent productive instead of dangerous.

Read →

How to stop an AI agent from "going rogue"

What "rogue" really means in practice, and the concrete controls that prevent it.

Read →

Hallucination & verification

The failure classes no LLM reviewer reliably catches — and the deterministic gates that do.

Your agent's data is lying to it

The most dangerous hallucination isn't made-up text — it's real code getting real APIs to return plausibly-wrong data, silently.

Read →

The verifier must fail differently from the verified

Why most agent self-review is theater, and the verification ladder that actually works.

Read →

Agent memory

Giving agents memory that survives sessions — and why the obvious answer isn't a vector database.

How to give an AI coding agent long-term memory (that it actually uses)

A file-based memory system, dogfooded on a real multi-week autonomous agent.

Read →

File-based memory vs a vector DB for AI coding agents

When plain files beat embeddings + retrieval — and when they don't.

Read →

Market-signal research

Honest, receipt-backed findings — including the ones that say "no edge."

We scored 100+ crypto analyst calls against real price. None beats a coin flip.

What a systematic, hindsight-clean accountability pass on public analyst calls actually shows.

Read →