Deciding better, with something that does not simply agree.
The thinking behind Five Peers: which reasoning methods hold up under evidence, which popular ideas do not, and why the memory matters more than the personalities. We name the sources, including where they are contested.
What an AI board of advisors actually is
The phrase covers two very different things. One is a costume change on a chatbot. The other is a decision process with roles, an order, and a record. Only the second one changes anything.
Read →The personal board of directors, and its problem
Assemble a handful of people who see different things, meet them regularly, bring them the hard calls. The idea is sound and old. What breaks is the logistics.
Read → The groundWhy it is lonely at the top, structurally
It is not a mood and it is not a flaw. It is what happens when you are the last stop for every question nobody else could settle, surrounded by people who all have an interest in the answer.
Read → The categoryDecision intelligence, minus the category jargon
The phrase describes software that helps people decide better. That is broad enough to be useless, so the useful distinction is what the tool actually does to the decision.
Read → EvidenceDevil’s advocacy works, and it costs you something
The evidence for arguing a decision out is solid and nearly forty years old. The evidence that it makes people like the process less is from the same studies, and it is the half that gets left out.
Read → EvidenceConfirmation bias, and why you cannot self-diagnose it
Knowing about a bias does very little to protect you from it. People shown direct evidence of their own bias continue to deny it, sincerely.
Read → EvidenceCognitive diversity: what the evidence supports, and what it does not
We build on this idea, so we have a reason to overstate it. Here is the honest version, including the peer-reviewed argument that the best-known formal result is wrong.
Read → EvidenceWhat forecasting tournaments teach about deciding
The largest study of whether judgement can be improved found that it can, cheaply, and told us exactly which procedures did it.
Read → EvidenceActively open-minded thinking, and why we measure it
Out of a long list of things you could measure about how someone decides, this one has the best ratio of predictive value to number of questions.
Read → EvidenceGroupthink, and the version that actually applies to you
The term is used to explain almost any bad group decision, which is a sign it explains very little. The specific mechanisms it points at are worth keeping.
Read → MethodThe pre-mortem: assume it already failed
One of the few debiasing techniques that is cheap, fast, and supported by a specific measured effect. The trick is grammatical: past tense, not conditional.
Read → MethodBase rates: how often does this actually work?
Before asking whether your plan will work, ask how often plans of this kind work. The answer is usually available, usually sobering, and usually ignored.
Read → MethodSecond-order thinking: and then what happens?
The first consequence is the one you intended. The second is the one that gets you, and it is usually visible in advance if anyone bothers to ask.
Read → MethodOne-way doors: which decisions deserve the agony
The useful sorting question is not how big a decision is. It is whether you can walk back through the door afterwards, and at what price.
Read → MethodCalibration: is your 80% actually 80%?
Being well calibrated means that things you are 80% sure about happen about 80% of the time. Most people are not, and almost nobody has checked.
Read → PracticeThe decision journal, and why almost nobody keeps one
Without a record written before the outcome, you cannot tell a good decision from a lucky one. Hindsight quietly rewrites what you believed, and it is very convincing.
Read → PracticeHow to run a meeting that actually decides
If the outcome is what the most senior person indicated in the first five minutes, you did not hold a decision meeting. You held a confirmation.
Read → The groundDecision fatigue, minus the overclaiming
The version you have heard is shakier than its popularity suggests. The underlying observation, that deciding repeatedly without support degrades what you produce, holds up on plainer grounds.
Read → PracticeUsing ChatGPT for real decisions: what it does badly
Not a criticism of the models. A description of what happens when a very capable, agreeable system meets a person who has already half decided.
Read → EvidenceWhy AI agrees with you, and when that is dangerous
Models trained on human approval learn that agreement is approved of. The behaviour is well documented and it scales with how clearly you signal what you want.
Read → The categoryThe AI executive coach: what it can and cannot be
Coaching is a relationship built on being known over time and being asked questions you would avoid. One of those is reproducible in software and the other is not.
Read → AlternativesPeer advisory groups: what they do well, and where they stop
Monthly rooms of non-competing leaders are genuinely valuable. They are also monthly, which is the entire problem when the decision is on Tuesday.
Read → AlternativesCoach, mentor, peer: three different jobs
A coach works on you. A mentor tells you what they did. A peer works the problem with you. The mismatch between what you needed and what you booked is usually the disappointment.
Read → DevelopersMCP explained, and what it does not give you
A useful, boring piece of plumbing that is regularly oversold. Worth understanding precisely, because the overselling is usually about portability.
Read → DevelopersPortable AI memory: mostly a claim about export
Memory is becoming the thing AI products compete on. Portability is becoming the thing they claim. One of these is easy to verify and the other is easy to assert.
Read →Five reasoning methods, chosen for the ways of thinking you do not use.
Compose my board