The 10x delivery method for AI teams

AI made coding fast. TenRail makes shipping fast.

Your agents write the code in minutes. Then the task sits for two weeks: unclear rules, a stuck decision, a blocked team, rework. TenRail removes that. You measure the result in your own repo, in two sprints.

$129 $99 Black Friday founder price, good until Dec 25

  • Works with Claude Code Cursor Codex Gemini
  • One-time payment · lifetime updates
  • 7-day money-back guarantee

The two weeks nobody budgets for.

A typical AI-assisted task, day by day.

The code took a day. Everything around it took nine.

Code is 20% of the task. That’s why AI gives you 1.2x.

Make that 20% five times faster and the task gets about 1.2x faster. Studies on AI coding measure 20 to 55%. To get 10x, you have to cut the other 80%.

Model factor

10.0x

10.00% of the original effort stays in the task

No single row gets you to 10x. Drop the approved spec and the model falls to 3.0x.

5%
5%
10%
10%
15%
10%
30%

TenRail removes the other 80%.

Each change moves work out of the task, before it starts or into an automatic check.

  • Without TenRail: The dev guesses the rule mid-taskWith TenRail: An approved spec with acceptance criteria arrives first
  • Without TenRail: An open question blocks the PRWith TenRail: Blocking decisions are settled before the task starts
  • Without TenRail: You wait on another team’s APIWith TenRail: Every team ships a contract and a fake on day one
  • Without TenRail: Bugs show up after the mergeWith TenRail: Tests from every criterion, gates in the PR, E2E run by the agent
  • Without TenRail: A deadline someone guessedWith TenRail: A factor you measure in two sprints
docs/specs/02-orders/ORD-01-place-order.md

ORD-01: Place orderApproved

Acceptance criteria

  1. AC-01Given an active customer and items in stockWhen the agent confirms the orderThen the order is created, stock is reserved and OrderPlaced.v1 is published
  2. AC-02Given an inactive customerWhen the agent tries to place the orderThen no order is created and the screen shows “Inactive customer”
  3. AC-03Given an item without enough stockWhen the agent confirmsThen nothing is reserved and the screen shows the missing item
  4. AC-04Given items priced in centsWhen the order is createdThen the total is the exact sum of quantity × price

Not a prompt pack. A method your CI can check.

Everything you need, installed in two commands.

Plus the tenrail CLI: one Python file, zero dependencies, runs in CI.

18

chapters of the method, in English, Portuguese and Spanish

22

templates: specs, ADRs, decision log, backlog, PR checklist

21

Claude Code skills, readable by any agent as Markdown

12

agents: spec writer, reviewer, decision broker, factor analyst and more

# copies the CLI, skills, agents and level N2 templates ./install.sh ~/code/shop N2 en "Shop"TenRail installed in ~/code/shop (level N2, language en).​# lint, trace and roadmap in one run python3 tools/tenrail.py checkcheck ok

Solo builder or fifty engineers. Same method.

Pick a level at install. It decides which files pay for themselves.

  1. N1SoloA short spec per feature, an AGENTS.md and tests that cite each criterion.
  2. N2Product teamFull specs, glossary, backlog, CI gates and E2E for every criterion with a UI.
  3. N3Platform & software housesLanes with contract and fake, collision rules and a roadmap with a critical path.
  4. N4Legacy replacementRules extracted with file and line, a golden master and cutover rehearsals.

Measured, not promised.

Run TenRail on your next tasks and measure your factor with tenrail measure. Not for you? One email within 7 days and you get every cent back. No questions.

7-day money-back guarantee

Built while planning the replacement of two legacy platforms: ~1,000 tables, ~180 scheduled jobs.

docs/measurement/cycle-2.mdtenrail measure

Cycle 2 report

KEEP
8.0xmeasured factor
10.0xplanned
7%rework
CycleTasksBaselineSpentFactorRework
128.0 d1.0 d8.0x10%
214.0 d0.5 d8.0x0%

Measured factor within 20% of plan.

Pricing

Pay once, no subscription, lifetime updates. Thirty days to ask for your money back.

Team

Up to 10 people, one invoice

$389$299

Get the Team plan
  • Everything in Individual, for up to 10 people
  • One charge and one invoice for the company
  • Lifetime updates for the whole team

Company

Unlimited people and a guided rollout

Let's talk

Email us
  • A licence for the whole company
  • Diagnosis, factor ADR and two measured sprints with your team
  • A report for leadership

Questions

Anything else, write to us and a person answers. Email us

Is 10x guaranteed?

No. 10x is the planning target. You measure the real factor over two sprints with rules fixed upfront. Below 4x, the plan is redone with the measured number.

Isn’t spec-first just waterfall?

No. Specs run one phase ahead, in waves. A task is detailed only when its lane is one or two tasks away, and a wrong spec is fixed in the same pull request.

Which AI tools does it work with?

Any agent that reads Markdown. Skills and agents install into Claude Code; Cursor, Codex and Gemini read the same AGENTS.md and skill files.

Which stacks?

Any. The CLI reads Markdown specs and a TOML backlog. Each gate maps to a tool you already use: pytest, Jest, Pest, Playwright, Stryker and others.

Does it work for one developer?

Yes. Level N1 keeps a short spec, an AGENTS.md and local gates. The measurement still runs and tells you when a higher level pays off.

What if it doesn’t work for us?

Email us within 7 days of your purchase and you get a full refund. No questions, no forms.