Uptick Systems
The reference deck behind the z21 proposal to Mila Ventures. Five bodies of practice, already tested at scale, packaged so your team and your agents can run them from day one.
Scroll
Where this comes from
Swipe the shelf to see all eight
Twenty-five years of other people's expensive lessons. Filled spines are what we install. Outlined ones wait on the shelf until the work calls for them.
The failure mode they all attack
A backlog is a pile of obligations nobody chose. An estimate is a guess wearing the costume of a commitment. Between them, a team stays busy for months with nothing shipped, and whoever spoke loudest last sets the roadmap.
Time is fixed. Scope is the variable.Shape Up
The map
Amazon supplies the first move, Basecamp the other four. Every idea enters at the front and earns its way forward. From here on the map stays at the top of the screen and the move you are in lights up. Depth is yours: Short gives one slide per move, Full adds the detail behind each.
Move 1 · Work backward · Working Backwards
Amazon writes the launch announcement first. If it reads as boring, so does the product, and that cost you one page.
A press release, plus the questions a customer and a skeptic would ask.
Six pages of prose, read in silence. Bullets hide weak thinking.
One leader carries the product, with nothing else competing.
Track what the team controls this week, not the revenue months later.
Move 1 · Work backward · Jobs to be Done
The PR-FAQ grew up around launches that were new to the world, so it quietly rewards an announcement that sounds momentous. Most good products are not moonshots. Jobs to be Done keeps the page honest: start from a switch that already happens in your market.
| Force | What it is |
|---|---|
| Push | The frustration with how they do this today |
| Pull | The appeal of the new way |
| Habit | The comfort of what they already have |
| Anxiety | What worries them about switching |
A switch happens when push and pull beat habit and anxiety. Write the PR-FAQ against that, and the press release stops being a wish.
Move 1 · Work backward · Product discovery
| Risk | The question |
|---|---|
| Value | Will anyone choose this over what they do today? |
| Usability | Can they work out how to use it without help? |
| Feasibility | Can we build it with the time, data, and skills on hand? |
| Viability | Does it work for the rest of the business, including sales and support? |
Cagan's point is about timing. These questions cost a conversation before the build and cost a quarter after launch.
Move 1 · Work backward · The idea maze
Dixon's picture of a good founder is somebody who has already walked the maze in their head. They know which paths others took, where those ended, and which early choice quietly rules out three later ones.
For an AI product, the early turns are the load-bearing ones. Which customer you serve first, where the data lives, how deep you sit inside their systems. Each of those forecloses options you may want in a year, so they get argued explicitly, on the record, before a line of code depends on them.
Move 2 · Shape · Shape Up
Turn a raw idea into a pitch somebody can answer yes or no to.
Ask how much time this is worth, then design to fit. Never how long it takes.
No mockups, no specs. The mechanism works end to end on paper; the edges stay open.
The pitch names what it will not do, so scope creep has nowhere to hide.
Output: one page with the problem, the solution, the rabbit holes, and the no-gos.
Move 3 · Bet · Shape Up
Each cycle, a few shaped pitches compete for one or two slots. The cap is on the cycle, not on your ideas. Keep the shelf as big as you like. A pitch that loses the table goes back on the shelf, uncommitted, and returns when somebody shapes it again. What dies is the obligation, not the idea.
Move 4 · Build · Shape Up
The team owns a project rather than a task list, and the first job is to make one thin path work end to end.
Move 4 · Build · Hill charts
Climbing means unknowns remain. Over the top means everything left is execution.
A marker that has not moved in two check-ins is the loudest signal on the chart. It means the scope hides an unsolved problem.
Move 4 · Build · The scope hammer
Scope grows because real work reveals real detail. So it gets hammered at the halfway mark and again near the end, before anyone has to ask.
Ask whether this beats what people have today.
Survivors are must-haves. Everything else is marked optional and dies quietly at the deadline. A cut version that ships whole beats a complete version that ships late.
Move 5 · Cool down · Shape Up
A deliberate stretch where nobody is on the hook. It is what keeps a team steering instead of permanently sprinting.
What you actually get
Each practice ships as a skill for Claude and Codex. The agent knows when to invoke it, so the process runs itself instead of living in a wiki nobody opens.
| Skill | Fires when | File in the repo |
|---|---|---|
| Working backward | A new product needs a PR-FAQ before anyone builds | ships with the engagement |
| Job interview | The PR-FAQ needs a real switch behind it | ships with the engagement |
| Shaping | Somebody brings a raw idea | mila-z21-hypothesis |
| Betting table | You ask what to build next | mila-z21-portfolio |
| Building | A bet is committed and the cycle starts | mila-z21-experiment |
| Hill chart | You ask how it is going | mila-z21-signal |
| Scope hammer | Halfway through, near the end, or when scope balloons | mila-z21-scope |
| Cool-down | Something ships and feedback arrives | mila-z21-debrief |
Install today: clone the repo and copy the skill folders into your agent's skills directory. Then ask the agent what to work on next.
git clone https://github.com/Uptick-Systems/z21 cp -r z21/skills/mila-z21-* ~/.claude/skills/
Show one, not describe one
One research direction from Lacuna went through shape and bet. Here is what came out, straight from the repo.
Appetite: small batch, one focused session. Problem: a web agent clicks the plausible next control and lands in a dead end, because it never simulated where the click leads. No-gos: no test-time training, no deep tree search.
Read the pitchFive questions answered on the record. Four pass, one flagged: no Mila author gravity in the direction. The session opens with a 20 minute go/no-go probe. If no drivable benchmark exists, the breaker fires before the cycle is spent.
Read the recordEight markdown files with YAML front matter. Drop them in your skills folder, bring a shelf of ideas, and run one through the loop.
Get the repoThe demo · 25 minutes
A row from the gap analysis, or a Lacuna direction dense with Mila authors.
The agent interrogates: appetite, breadboard, rabbit holes, no-gos. Out comes a framed hypothesis.
Five funding questions. The agent takes a position and writes the kick-off.
One vertical slice of the riskiest mechanism, end to end.
Hill status, one staged scope refusal, the dossier on screen.
In practice
A product with a long arc and real unknowns is exactly the shape this rhythm handles well. Here is how the first months usually run.
The PR-FAQ decides which customer you serve first and what ships on day one. The idea maze names the turns that close doors.
Shaped pitches and capped bets, riskiest mechanism first, the four risks answered before each build.
When the product is agentic, your team running the process is the first proof it works.
Next step
I need whatever product material you already have and two working sessions in week one. I set up everything else.
Addendum · Also on the shelf
These do not run every cycle. Each one earns its place when a particular problem shows up, and I bring it then rather than installing it up front.
A weekly interview habit and the opportunity solution tree. For when you need to tell a pattern from one loud account.
Judge the product by what the user becomes good at. For when the pitch drifts toward what the software does.
Plot each component by how commoditized it is: what to build, what to rent. The sharpest idea-maze tool, and it takes practice.
Addendum · Considered and set aside
| Method | Reason |
|---|---|
| Lean Startup | A decade of loose use has worn the MVP idea smooth. Shape Up's appetite does the same job with a harder edge. |
| Double Diamond | A design agency frame. It describes the shape of work without saying who decides or when to stop. |
| SAFe and its kin | Built for coordinating dozens of teams. A small team has the opposite problem. |
| OKRs | Fine for direction, weak on how work gets picked. If you already run them, the cycles slot underneath without a fight. |
Naming these matters, because a process nobody argued about is a process nobody owns.
Read the originals