Pocket Tennis Team Season Plan
Pocket Tennis Team · release and development timeline · Oct 2026 – Jun 2027

From preview to a team people pay for

The full plan for Pocket Tennis Team: ship a public preview in 11 days, grow it with a friends beta, make it a strong portfolio piece, then decide with real evidence whether to turn it into a paid product in March 2027.

Today: Thu, Oct 15, 2026 Capacity: about 10 h a week (you), with AI agents doing most of the build and testing License: PolyForm Noncommercial; commercial rights stay with you Dates are targets; holidays have buffer built in

Season chart

Five tracks run in parallel. Solid bars are planned; dashed ones only happen if the March decision is "go". Scroll sideways on a phone.

Releases Development Testing and feedback Portfolio and sharing Business Only if "go" in March

The next four weeks

Bold items need you; the rest the agents and I can do. Checked items are already done.

Oct 15 – 18 · this week

Get it saved and online

  • Switch to the non-commercial license
  • Clean up for public view, portfolio README
  • Check your employment contract for side-project rules
  • Share your GitHub username; approve first commit
  • Create the repo, push, add topics and description
  • Turn on GitHub Pages
Oct 19 – 25

Prove it on real Macs

  • Live demo on GitHub Pages (works without Claude)
  • Screenshots and a short GIF for the README
  • Real-mic talk-mode test; tell me how it feels
  • Fresh install on a second Mac (friend or spare)
  • Auto-open check in the CLI, Desktop Code tab and Cowork
  • Fix anything that blocks a first-time user
Ship Mon, Oct 26 · v0.2.0 Preview
Oct 26 – Nov 1

Friends beta starts

  • Invite 5–8 testers: 3–4 adult players, 1–2 parents
  • Beta kit: install guide, demo link, 2-minute feedback form
  • Build: plan review appears instantly (draft while setup runs)
  • Case study page live (from the Blueprint)
  • Add the project to your resume and portfolio
Nov 2 – 8

First fixes, first post

  • v0.2.1 patch from beta feedback
  • Build: lighter background updates, daily cost cap
  • 60-second demo video
  • LinkedIn or blog post: "how I built an AI tennis team"
Ship Mon, Nov 9 · v0.2.1

Phases and versions

Each phase ends in a release with clear exit criteria, so it's always obvious what "done" means.

PhaseDatesVersionScopeDone when
0 · PreviewOct 15 – Oct 26v0.2.0-preview
  • License, cleanup, portfolio README
  • GitHub repo and Pages demo
  • Real-device test pass and blocker fixes
Installs from GitHub on 2 Macs; demo link works without Claude
1 · Launch-ready betaOct 27 – Nov 22v0.2.1 – 0.2.4 (weekly)
  • Instant plan review
  • Lighter bookkeeping (about 30% of today's cost)
  • Daily cost cap; friendly error recovery
  • Case study, demo video, first post
5+ active testers; no data-loss bugs; reply cost under half of today
2 · Feels like a real teamNov 23 – Jan 18v0.3.0
  • AI understanding replaces keyword rules
  • Streaming replies (coach speaks in about 5 s)
  • On-device Whisper recognition
  • Player Scout and Fitness Coach
  • Holiday buffer Dec 21 – Jan 3
Persona retest 4.5+ on every stage; coach starts talking within 5 s
2b · DirectoryJan 25 – Feb 21v0.3.x
  • Privacy note and listing assets
  • Submit to Anthropic's plugin directory
  • Review time isn't in our control
Submitted; feedback from the review addressed
3 · Family and on the goJan 19 – Mar 28v0.4.0
  • Knowledge library part 2, Sunday check-in, match tools
  • Phone voice memos, then phone access
  • Nutrition (adults), Google Calendar sync
Parents use it courtside; phone notes reach the team without a laptop
4 · Product (if go)Mar 30 – Jun 21v1.0 hosted beta
  • Hosted version: accounts, Claude API backend, phone-first
  • Pricing, privacy policy, minors compliance
  • Coach and academy pilot
10+ paying users or a signed academy pilot

Decision gates

Four checkpoints where the evidence decides the next step, not enthusiasm.

G1 · Preview shipped

Mon, Oct 26
  • Install from GitHub works on 2 Macs
  • Live demo runs without Claude
  • README and case study are ready to show
If not: slip one week; never ship a broken first run.

G2 · Beta health

Sun, Nov 22
  • 5+ testers active
  • 60% log a second note within 7 days
  • Average reply under 20 s; no lost notes
If not: interview the drop-outs and fix the first-week experience before v0.3 features.

G3 · v0.3 quality

Mon, Jan 18
  • Persona retest 4.5+ everywhere
  • Coach starts speaking within 5 s
  • Ready for directory review
If not: keep polishing; directory submission waits.

G4 · Profit decision

Mon, Mar 1
  • 40% of testers still active at week 8
  • 3 of 10 interviewees would pay a monthly fee (test about $8–15)
  • One clear segment: adult players, parents, or coaches
  • Per-user AI cost fits under the price
If no-go: keep it as an open portfolio project with small releases. Nothing is wasted.

Risks to watch

RiskWhy it mattersWhat we do
AI cost per messageEach message is two AI runs; background bookkeeping cost about $0.90 in one testPhase 1 cost work and a daily cap; a hosted price must cover it
Your side-project rulesSome employment contracts claim side projectsCheck before making the repo public (this week)
Solo timeAbout 10 h a week; life and tennis happenWeekly patches stay small; holiday buffer; agents do the heavy lifting
Platform changesClaude plugin features are still evolvingKeep the server and skills simple; follow release notes
Minors and privacyJunior players and health notesParent-operated only; local data; a legal review before any hosted version
Directory review timingNot in our controlGrow through GitHub and friends first; the directory is a bonus