https://airunning.co.uk/Recreational runners training for a 5K to marathon — typically already following an app like Runna, Garmin Coach or a ChatGPT-built plan — who can read their own Strava and watch data but need help judging whether an AI-generated plan is progressing them safely.https://airunning.co.uk/blog/001-the-four-numbers-worth-checking-every-sunday-night/https://airunning.co.uk/blog/001-the-four-numbers-worth-checking-every-sunday-night/Most runners drown in watch metrics; argues that weekly volume, weekly volume trend, hard-session count and a subjective fatigue note are enough to judge whether any plan is progressing you safely, and everything else is optional detail.https://airunning.co.uk/blog/002-what-the-ai-in-your-running-app-is-actually-doing/https://airunning.co.uk/blog/002-what-the-ai-in-your-running-app-is-actually-doing/Separates genuine adaptive modelling from rule-based templates with a marketing label, and gives readers three questions to ask of any app to work out which one they've bought.https://airunning.co.uk/blog/003-acute-chronic-workload-ratio-explained-without-the-jargon/https://airunning.co.uk/blog/003-acute-chronic-workload-ratio-explained-without-the-jargon/Defines ACWR as 'this week versus your recent normal' and argues its value is as a direction-of-travel check, not a precise injury predictor with a magic 1.5 cliff.https://airunning.co.uk/blog/004-garmin-training-readiness-what-actually-goes-into-the-score/https://airunning.co.uk/blog/004-garmin-training-readiness-what-actually-goes-into-the-score/Breaks the composite score into its inputs — sleep, recovery time, HRV status, acute load, stress — and shows which input is usually driving a surprising number so you can judge it rather than obey it.https://airunning.co.uk/blog/005-your-first-chatgpt-training-plan-prompt-line-by-line/https://airunning.co.uk/blog/005-your-first-chatgpt-training-plan-prompt-line-by-line/Builds one complete prompt from scratch, explaining why each element (current weekly mileage, recent race, constraints, injury history, goal date) changes the output, with before-and-after plans showing the difference.https://airunning.co.uk/blog/006-what-ai-form-analysis-can-see-from-a-phone-video-and-what-it/https://airunning.co.uk/blog/006-what-ai-form-analysis-can-see-from-a-phone-video-and-what-it/Sets honest expectations: 2D video apps can estimate cadence, overstride and trunk lean reasonably, but cannot measure forces, rotation or anything that maps cleanly onto injury risk.https://airunning.co.uk/blog/007-the-first-two-weeks-back-ground-rules-before-any-plan/https://airunning.co.uk/blog/007-the-first-two-weeks-back-ground-rules-before-any-plan/Argues that no AI plan should be started until you can pass simple load tests pain-free, and gives the criteria-based gates that matter more than any calendar date an app offers.https://airunning.co.uk/blog/008-adaptive-or-just-rescheduled-how-coaching-apps-handle-a-miss/https://airunning.co.uk/blog/008-adaptive-or-just-rescheduled-how-coaching-apps-handle-a-miss/Tests what several apps do when you skip a key session or a whole week — some genuinely rebuild the block, most just shuffle tiles — and explains why that distinction decides how much you can trust the plan.https://airunning.co.uk/blog/009-is-the-10-rule-real-or-just-convenient/https://airunning.co.uk/blog/009-is-the-10-rule-real-or-just-convenient/Compares the flat 10% weekly-increase heuristic against a rolling-average approach, showing that 10% of a small base is trivially safe while 10% of a big base can be a genuine spike.https://airunning.co.uk/blog/010-your-hrv-dropped-here-s-what-that-does-and-doesn-t-mean/https://airunning.co.uk/blog/010-your-hrv-dropped-here-s-what-that-does-and-doesn-t-mean/Distinguishes a single overnight dip (usually noise, alcohol or a late meal) from a multi-day downward trend against baseline, and only treats the second as a reason to change the week.https://airunning.co.uk/blog/011-feeding-the-model-your-strava-history-without-feeding-it-jun/https://airunning.co.uk/blog/011-feeding-the-model-your-strava-history-without-feeding-it-jun/Shows how to export and condense twelve weeks of running into a short structured summary the model can reason over, and why pasting raw activity dumps produces worse plans than a tidy ten-line table.https://airunning.co.uk/blog/012-cadence-is-the-one-form-metric-worth-changing-first/https://airunning.co.uk/blog/012-cadence-is-the-one-form-metric-worth-changing-first/Debunks 180 as a universal target, then argues that a modest 5% cadence increase is the single form change with decent evidence behind it and the easiest to self-coach with watch feedback.https://airunning.co.uk/blog/013-miles-minutes-or-load-points-which-unit-should-you-track/https://airunning.co.uk/blog/013-miles-minutes-or-load-points-which-unit-should-you-track/Compares distance, duration, sRPE and Garmin/TrainingPeaks load scores, and recommends duration-based load for most recreational runners because it doesn't punish hills, heat or easy weeks.https://airunning.co.uk/blog/014-why-your-app-s-easy-pace-feels-hard/https://airunning.co.uk/blog/014-why-your-app-s-easy-pace-feels-hard/Explains how apps derive paces from a race result or VO2max estimate, why that overstates fitness for under-trained or heat-affected runners, and how to sanity-check prescribed paces against your own heart rate data.https://airunning.co.uk/blog/015-resting-heart-rate-is-the-slow-boring-metric-that-actually-w/https://airunning.co.uk/blog/015-resting-heart-rate-is-the-slow-boring-metric-that-actually-w/Argues RHR trend over weeks is a more reliable fitness-and-fatigue signal for recreational runners than any daily readiness score, because it's simple, stable and hard to game.https://airunning.co.uk/blog/016-five-things-ai-gets-wrong-when-it-writes-a-running-plan/https://airunning.co.uk/blog/016-five-things-ai-gets-wrong-when-it-writes-a-running-plan/Catalogues the recurring failure modes — impossible mileage jumps, too many quality days, no down weeks, invented paces, taper that starts too late — with a checklist for spotting each in a generated plan.https://airunning.co.uk/blog/017-the-pain-rules-when-to-keep-running-and-when-to-stop/https://airunning.co.uk/blog/017-the-pain-rules-when-to-keep-running-and-when-to-stop/Sets out the traffic-light framework — pain level during the run, and how it behaves in the 24 hours after — as the decision rule that overrides whatever the comeback plan says.https://airunning.co.uk/blog/018-down-weeks-how-much-to-drop-and-how-often/https://airunning.co.uk/blog/018-down-weeks-how-much-to-drop-and-how-often/Argues for fatigue-triggered rather than calendar-fixed recovery weeks, and shows how to check whether your AI plan's 3:1 pattern actually reduces load or just reshuffles it.https://airunning.co.uk/blog/019-overstriding-what-it-really-looks-like-on-video/https://airunning.co.uk/blog/019-overstriding-what-it-really-looks-like-on-video/Teaches readers to judge foot-strike position relative to the knee and centre of mass from their own slow-motion footage, rather than trusting an app's red-flag label.https://airunning.co.uk/blog/020-why-garmin-keeps-telling-you-your-training-is-unproductive/https://airunning.co.uk/blog/020-why-garmin-keeps-telling-you-your-training-is-unproductive/Explains the load-versus-VO2max-trend logic behind the label, why it misfires during base building, heat and taper, and the two cases where it genuinely deserves attention.https://airunning.co.uk/blog/021-which-app-would-you-trust-with-a-16-week-marathon-build/https://airunning.co.uk/blog/021-which-app-would-you-trust-with-a-16-week-marathon-build/Judges the major AI coaching apps on the things that decide a marathon — long-run progression, quality-session balance, taper design and how they handle a missed fortnight — rather than on UI polish.https://airunning.co.uk/blog/022-making-ai-build-a-long-run-progression-that-holds-up/https://airunning.co.uk/blog/022-making-ai-build-a-long-run-progression-that-holds-up/Gives the constraints to write into your prompt — cap as a share of weekly volume, alternate step-back weeks, add quality inside the long run late — and shows an unconstrained plan failing against them.https://airunning.co.uk/blog/023-audit-your-plan-s-load-curve-before-you-run-a-step-of-it/https://airunning.co.uk/blog/023-audit-your-plan-s-load-curve-before-you-run-a-step-of-it/A practical walkthrough of plotting a generated 16-week plan's weekly load and looking for spikes, plateaus and a taper that starts on time — the single highest-value hour you can spend on an AI plan.https://airunning.co.uk/blog/024-body-battery-whoop-recovery-and-the-number-that-disagrees-wi/https://airunning.co.uk/blog/024-body-battery-whoop-recovery-and-the-number-that-disagrees-wi/Compares how different recovery scores are constructed and argues that when the score and your legs disagree, the score is the one that should lose — with the exception cases spelled out.https://airunning.co.uk/blog/025-ground-contact-time-and-vertical-oscillation-useful-or-noise/https://airunning.co.uk/blog/025-ground-contact-time-and-vertical-oscillation-useful-or-noise/Shows these watch metrics track speed more than technique, so they're only informative when compared against yourself at matched pace — and mostly useless as cross-runner benchmarks.https://airunning.co.uk/blog/026-rebuilding-chronic-load-from-zero-without-overshooting/https://airunning.co.uk/blog/026-rebuilding-chronic-load-from-zero-without-overshooting/Explains why ratio-based load rules break down at low baselines and proposes absolute weekly increments for the first six weeks back instead, with worked numbers.https://airunning.co.uk/blog/027-garmin-daily-suggested-workouts-three-weeks-of-following-the/https://airunning.co.uk/blog/027-garmin-daily-suggested-workouts-three-weeks-of-following-the/A logged test of obeying the watch's daily suggestions, showing where the reactive day-to-day model conflicts with the structure a race build actually needs.