Roadmap

What has shipped, what we are working on, and what comes next. One rule sets the format: dates appear only where something has already shipped. Everything else carries a status, not a deadline. Anything that touches the physics goes through pre-registered validation on the methodology page before it lands, and a safety-related fix never waits for a calendar date.

Shipped — live today

• The internal ballistics engine. A physics simulation that steps through the shot in half-microsecond increments, keeping an energy ledger: burning powder adds energy; bullet travel, heat into the steel, rifling friction and gas leak past the projectile subtract it. Pressure follows from the energy still in the gas. live today
• Charge ladders with uncertainty bands. The work-up ladder shows how predicted pressure moves with charge weight, and every rung carries an uncertainty band. Safety verdicts are judged at the band's conservative top edge, never the displayed value — above the top of the range the verdict stops asserting safety rather than guessing. live today
• Trajectory drop tables. Shipped 2026-11-01. The engine predicts the muzzle velocity; a point-mass external model (drag + gravity) turns that into a drop table. Free tier shows the table out to 300 yd; PRO gets the full range. Cross-checked against an independent published solver before it shipped. shipped 2026-11-01
• Load logbook. Save loads to your device, export them as CSV or JSON, and print a load sheet that puts the predicted velocity and pressure band next to your work-up notes. live today
• Printable range sheets. Take the numbers to the bench on paper: a clean printed sheet with the table your load needs and the safety notes that travel with it. live today
• MCP server for AI agents — live beta. AI agents (Claude Code, Cursor, and any MCP client) can run loads through the engine directly: simulations, charge ladders, trajectory tables, catalog lookups. Same safety verdicts, same published error figures, no separate claims. Free beta keys on the MCP page; 100 engine calls a day per key while the beta is tuned. shipped 2026-11-02
• Published accuracy with provenance. The error figures are not a marketing line — they are computed from the validation corpus on every deploy, with the source data spelled out, and they render on this page from the served artifact rather than being typed in: published in full on the methodology page (full disclosure). live today

In development

• Chronograph calibration. in development
  Enter the velocities your own chronograph measured, and the model tunes to your rifle, your lot of powder, your barrel. This is the flagship feature: published data is shot from proof barrels, and the honest way to close that gap is to calibrate against the one barrel you actually own. The design exists; it ships when it clears validation, not before.
• Efficiency-descriptor fix for the high-reading tail. in development
  The worst error regime is loads the model reads high. The pre-registered mechanism study is complete and pinned the cause: efficiency collapse tied to charge mass relative to bullet mass — the case-fill framing did not survive the study. The fix is an efficiency-descriptor term; it will run the same before-and-after validation on the methodology page as every engine change.

Coming — planned

• Pistol cartridge coverage. planned
  Coverage is rifle-focused today, and we say so everywhere rather than quietly extrapolate handgun loads from rifle data. Pistol cartridges are next; interior ballistics in near-empty cases is the hardest regime in the field, so this lands only when it clears the same accuracy floor as the rifle catalog.
• The accuracy paper. planned
  A full methodology write-up: corpus construction, error distributions, where the model errs and why, and what the figures are good for. Everything on the methodology page today is true; the paper makes it complete.

Plans, not promises. Every item above is a plan, not a promise. The accuracy page discipline applies to claims, and we hold schedules to the same standard: we would rather under-promise and ship than put a date on something that then slips. If a change would move the published figures, the figures are published before and after, and a safety-related fix jumps every queue.