Research
Questions first. Products second.
We research how AI and on-chain software can be trusted with real work and real money. Three areas guide everything we build.
Research · AI
AI
Agents you can hand real work to.
- Agentic software engineering
- Grounding and verification
- Human supervision
- Evaluation
Blueprint · Superteam · Forge · Superfam
Research · Crypto
Crypto
Infrastructure for money that runs on-chain.
- Market microstructure
- Autonomous risk
- Self-custody treasury
- Validator operations
Hayek · Vector
Research · d/acc
d/acc
Accelerate the technology that keeps people free and safe.
- Defense over offense
- Self-sovereign tools
- Decentralized infrastructure
- Open verification
Superfam · Hayek
Latest writing
October 6, 2026 · 8 min read
The most valuable code in a trading system is the code that says no
Modern tools make thousands of backtests cheap. Statistics says that's exactly the problem.
Every backtest you run makes the next one less believable. The durable edge is in the rules that refuse trades.
Read →
September 29, 2026 · 8 min read
Models are rented. Memory should be owned.
The best model changes every few months. The memory you build using it shouldn't be stuck with whoever was ahead when you started.
If you can't take your memory to another model, it isn't your memory.
Read →
September 17, 2026 · 8 min read
Give AI to your newest hire first
Most businesses roll AI out to their best people. The evidence says the biggest payoff is at the bottom of the learning curve.
AI is a ramp, not a raise: it helps your newest people most, so give it to them first.
Read →
September 2, 2026 · 9 min read
AI tutoring is a motivation problem, not an intelligence problem
Tutors that give answers make students worse. Tutors that refuse get ignored. Smarter models fix neither.
AI tutoring is a motivation problem wearing an intelligence costume.
Read →
August 21, 2026 · 8 min read
For agentic work, the note is the deliverable
When agents do work nobody watched, the account of the work becomes more valuable than the work itself.
If a task can't be redone from its closing note, it isn't done.
Read →
August 3, 2026 · 9 min read
The best agent needs the model less every week
Agents are judged by how smart their decisions are. We think they should be judged by how few decisions they still need a model to make.
Measure an agent by how small its judgment surface gets, not by how smart its judgments are.
Read →
July 16, 2026 · 9 min read
Save less: the case against total recall
Personal AI is racing to remember everything. The evidence from cognitive science and from language-model retrieval points the other way.
A memory system should be judged by what it lets you forget, not by what it keeps.
Read →