technology · general

Fair Winds: how we made starting easier and progress visible

Bradley Hunt ·
program generation AI coaching training programs product update benchmarks personal records

Fair Winds is the release where Pelaris gets out of your way.

Fair Winds takes its name from the sailor’s farewell, “fair winds and following seas”: conditions that carry you forward instead of making you fight for every metre. That is the whole theme. Less friction between you and your next session, and the progress you have already banked made visible instead of buried.

The uncomfortable finding behind this release: most people who generate a training program never complete a session. Not because the programs are wrong, but because too much stood between “I want to train” and training, and too little showed people what they were building. Fair Winds attacks both ends. Here’s what changed and why.

Start from where you are

The Today screen is the first thing you see, and until now its “Create your program” card bounced you to a different tab to actually do it. One tap became three, and every extra tap at that moment loses people.

In place, no detour. The create card now opens the program flow right where you are. The same applies to rest days: tap any rest day in your schedule and you can add a session there, a quick swim, run or ride, without hunting through menus. Training plans should bend to your week, not the other way around.

This also fixes a real bug we found while building it: sessions added from the schedule viewer could stamp the wrong date and land on the wrong day. That is gone.

Your program arrives as a draft, not a fait accompli

Program generation used to finish by silently enrolling you. Done, live, hope you like it. That worked, but it put the machine’s confidence ahead of yours.

Review, then activate. Generation now lands as a draft. You get a “Ready to review” card, you look over the weeks, and you activate it when you’re satisfied. If you close the app mid-generation, the draft is waiting when you come back. Nothing goes live without your say-so.

Under the surface this rebuilt how enrolment works so that the failure modes we used to patch around, like a program enrolling twice, are designed out rather than caught late. You should never notice that part, which is the point.

The generator finally reads what you write

This one deserves the plainest possible telling. The program creation screen has a text box for context: injuries, preferences, the routine you already run. People used it. And for months, what you typed never reached the AI that built your program.

Fixed, properly. Your notes now go to the generator, below the safety rules, so “I’m rehabbing a knee, keep squats light” actually shapes the plan. If you asked Pelaris to mirror an existing routine, the parsed version of that routine now anchors week one the way it always should have.

We could have quietly fixed this and said nothing. But the promise of an AI coach is that it listens, and for a while ours didn’t. If you typed something that mattered before this release, regenerate. It will be read this time.

Generation that tells the truth

Waiting on an AI to build a multi-week program is a trust exercise, and we were losing it in two ways: waits that looked hung when they weren’t, and retries that could hand you two programs.

Honest progress, safe retries. The generation screen now shows clearer stages with a soft time estimate, so a slow patch reads as working rather than wedged. If a wait genuinely stalls, you’re routed to retry instead of being left staring. And retry itself is now idempotent: if the server actually finished while your connection wobbled, retrying adopts that finished program instead of generating a second one. Stale jobs get cleaned up server-side so nothing sits “in progress” forever.

The wins you already earned, on screen

Pelaris has been quietly tracking personal records and benchmarks for a long time. Too quietly. The data was there; the payoff wasn’t.

Front and centre. Your PR and benchmark timeline now leads the Horizon view, so “how am I doing?” is answerable in about five seconds. Sections below it collapse and remember your preference.

Both halves of hybrid. The benchmark catalogue expands past the barbell: alongside the key lifts, Pelaris now tracks endurance PRs across swim, run and ride distances, from a 100m sprint to the marathon. Timed efforts display as proper mm:ss splits instead of raw seconds, and faster correctly counts as better.

If you train for a 5k and a deadlift in the same week, and plenty of you do, both now show up as progress. That combination is the athlete Pelaris is built for.

The analytics other apps charge for

Fair Winds also rounds out the analytics suite that landed alongside it: Strength Progress with estimated one-rep-max trends, a Weekly Report, the PR Timeline, a Hybrid Scoreboard that puts your lifting and endurance side by side, and CSV export of your complete history. The trackers we get compared to paywall almost exactly this list, which is why the section carries a plain caption: every stat and chart, always free. Tracking and understanding your own training should not be a subscription tier. If you want the reasoning, we wrote about why the loop from your data back into your next session matters more than the log itself, and there’s a full comparison against the popular trackers.

Horizon itself now opens as a clean, scannable index: each section starts collapsed with a one-line hint of what’s inside, and your expand choices are remembered across devices.

Ask the coach for a chart

The newest piece ships right at the release line: the coach can now answer with a chart. Ask “show my bench trend” or “how has my weekly volume moved” and the reply renders the actual graph under the coach’s caption, drawn from your real training data.

Bounded on purpose. The AI picks which chart and which data series from a fixed, code-owned vocabulary, and the app validates and renders it with the same chart kit the analytics tab uses. The model never draws free-form output, so a chart in a coach reply is always your data through our renderers. If a request can’t be answered from that vocabulary, the coach says so in plain text rather than improvising. Start narrow, render honestly, widen the vocabulary as it earns trust.

A proper finish line

Finishing a session used to end with a tick. Functional, and completely flat as a moment.

A debrief, every time. Finish a session and Pelaris writes you a short, personal coaching summary: what you did, what it means, what it sets up. It was technically possible before but hidden behind conditions most sessions never met. Now it’s the reward for showing up, every time.

The small stuff that adds up

A release named for smooth passage should also clear the barnacles:

  • Days with two sessions now show a count badge in the weekly calendar, so doubles are no longer collapsed into a single dot
  • The coach can now skip or delete a planned session that hasn’t started yet, instead of pretending it can’t find it
  • Synced activities from Strava now expose their per-lap splits (pace, heart rate, cadence, power, swim strokes) to the coach and connected AI tools, so feedback can reference the lap you actually swam or ran
  • Editing a session through the coach now resolves exercises against the library the same way creating one does, keeping your history and PRs stitched to the right movements

What’s next

Fair Winds is the foundation for a run of work aimed at the same target: making the gap between “generated a program” and “trained this week” as small as possible. Next up is a deeper body analysis that interprets your measurements through your sport and goals rather than generic norms, and a more conversational intake so starting feels less like a form. Same course, following seas.

What this means for your training

  • Fewer taps to start. Create from Today, add sessions to rest days directly, and get going
  • You approve your program. Generation produces a draft; nothing goes live until you activate it
  • Your context counts. Notes and existing routines you provide now genuinely shape what gets built
  • Progress is visible. PRs, benchmarks and a full analytics suite across lifting and endurance, every stat and chart free
  • The coach can show you. Ask for a trend and the reply comes back with the actual chart, drawn from your data
  • Finishing feels finished. Every completed session ends with a personal coaching summary, not just a tick