Demos Not Memos
The first 10% of every project is now free — so build something to react to instead of writing documents.
- Difficulty
- Easy
- Time to result
- ~days to results
- Steps
- 3
- Confidence
- 86%
With AI, the first 10% of any project is now free: it takes almost no effort to build a janky first version. Schoening's operating rule, carried from GitHub, is 'demos not memos' — give people something concrete to react to rather than a PRD or memo. This bakes iteration into the process early, replaces waterfall planning, and lets you cheaply explore many paths in parallel (increasing 'shots on goal'). The hard part hasn't moved — the last 10% is still 90% of the work — but the exploration phase is now nearly frictionless.
Origin
Schoening credits the 'demos not memos' and 'give me something to react to' practices to GitHub product reviews; the AI-era update is that producing the demo is now nearly free.
Core principles
- 01The first 10% of every project is now free
- 02Demos beat memos — give people something concrete to react to
- 03The last 10% is still actually 90% — the hard part hasn't moved
- 04Cheap exploration lets you run many parallel paths
- 05Increase 'shots on goal' rather than perfecting a single plan on paper
How to run it
- 1
Skip the PRD, build the janky version
Instead of writing a PRD, produce the janky demo of what you think should be built. If you must write, write the change log or the blog post a user would read — the artifact that lets people react.
Pro tip 'Give me something to react to' — a rough working version elicits far better feedback than a document.
Watch out Don't over-invest in the first 10% precisely because it's now cheap; the value is speed to something reactable.
- 2
Explore many paths in parallel
Because exploration is cheap, send off multiple agents to explore several different approaches at once, then see which was right. This increases shots on goal.
Pro tip Treat parallel exploration as the default; waterfall single-path planning is now wasteful.
- 3
Respect that the last 10% is still 90%
The cheap first 90% can create false confidence. The remaining work — making it reliable for millions — is the real engineering and hasn't gotten easier.
Pro tip Budget most of your effort for the final stretch, not the exciting free opening.
Watch out Software quality has not risen with quantity — a cheap demo is not a shippable product.
In the wild
At GitHub, product reviews ran on 'demos not memos' and 'give me something to react to' — if you were going to write a PR, you'd write the change log or the blog post a user would read, so people had something concrete to respond to.
→ Built iteration into the product process early and avoided waterfall memo-writing.
Schoening describes now being able to say 'I'm going to send off 10 agents to explore 10 different things and then see if I was right,' making it cheap to explore many product directions at once.
→ Iteration is built in much earlier; exploration cost collapses and shots on goal rise.
Common mistakes
Writing memos when a demo would do
Producing PRDs and memos for things that could be shown as a rough working version wastes the now-free first 10% and gives people worse material to react to.
Mistaking the free 90% for done
Because the first version is nearly free, teams underestimate the last 10%, which is still 90% of the real work — reliability at scale. Quantity of software has risen without quality.
Is it for you?
Best for
Product teams and founders in the exploration phase who want to replace planning documents with fast, reactable prototypes
Not ideal for
The reliability-critical final stretch of a product, where cheap demos give false confidence and rigorous engineering is required
From the transcript
“The first 10% of every project are now free.”
“demos not memos”
“if you're going to write a PR just write the change log or the blog post that a user would have would read”
“I'm going to send off 10 agents to explore 10 different things”
From the episode
Why cultivating agency matters more than cultivating skills in the AI era
Max Schoening (Head of Product, Notion)