AirJelly — a free, macOS-only desktop agent that observes your work locally and auto-captures tasks from meetings and messages — is worth trialing for the commitments you never write down. Judged by Sir John Whitmore's Coaching for Performance equation, performance equals potential minus interference, it scores 4/5 on interference reduction but only 2/5 on building responsibility.
Every founder I coach has a version of the same leak: the commitment made out loud on a Tuesday call that never became a task, rediscovered three weeks later as an apology email. Task managers do not fix this, because the tasks that hurt you are precisely the ones you never wrote down. So when a new tool launched with exactly that pitch, I paid attention.
AirJelly is an always-on macOS agent, freshly launched and currently sitting on Product Hunt's productivity charts, that bills itself as the world's first context-aware proactive agent. It watches your work across Slack, Zoom, Docs, and Calendar, builds a searchable timeline of everything it observed, auto-captures tasks with due dates from your conversations, produces an end-of-day report of what happened and what is pending, and surfaces briefs and follow-ups before you ask. Two facts define the current offer. First, everything runs locally — no cloud sync, no training on your data, per the company's own privacy claims — which is the only acceptable architecture for a tool whose job is watching everything you do. Second, it is entirely free right now, with no published paid plan; MakerStack's review rates it 6.5/10 and notes that Motion and Reclaim charge roughly $10 to $35 a month for adjacent AI scheduling, so the honest price of AirJelly today is your Mac's spare compute and your trust.
A tool this ambitious needs a harder test than a feature list, so I used the oldest one in the coaching literature. In Coaching for Performance, Sir John Whitmore builds everything on one equation — performance equals potential minus interference — and two pillars: awareness, seeing your reality clearly, and responsibility, owning the next action yourself. His career-long warning was that telling people what to do creates dependence and suppresses both. A proactive agent is, by design, a telling machine. Here is where AirJelly lands against that standard after a week of testing:
| Whitmore criterion | Score | Why |
|---|---|---|
| Interference reduction | 4/5 | Catching unwritten commitments removes the fear-of-forgetting tax that hums under a founder's whole day |
| Awareness | 3/5 | The daily report and timeline genuinely show you your real day; passive capture can also dull your own noticing |
| Responsibility | 2/5 | Proactive alerts decide what matters for you — Whitmore's dependence problem, automated |
| Trust cost | 4/5 | Local-only processing, nothing uploaded; the trade is your machine carries the compute |
| Production maturity | 2/5 | Early product, macOS only, free with no visible business model yet |
The split matters more than the average. On interference, AirJelly earns its keep the first time it converts a meeting aside into a tracked task you would have dropped. The end-of-day report is quietly the best feature by Whitmore's standard, because it raises awareness without commanding anything — a five-second honest answer to what actually got done. But the proactive layer is where the category lives or dies, and MakerStack's line matches my experience: when it works it feels like an assistant who reads ahead, and when it misfires it is an interruption you did not ask for — which is interference, manufactured by the tool that promised to remove it. The practical costs are real too: continuous local inference means battery drain and fan noise on older hardware (recent Apple Silicon handles it fine), and a free tier with no stated pricing will change, so do not wire your whole operating rhythm into it yet.
My verdict after the test: trial it for capture, refuse it as a decider. Let it catch what you drop and report what you did; keep the choice of what tomorrow's priority is as a decision you make yourself, out loud, every morning. The founders who struggle most in coaching are rarely short of reminders — they are short of ownership, and no agent, however context-aware, can be responsible on your behalf. A tool that remembers everything is useful. A tool you let think for you is a slow leak in exactly the capacity that makes you fundable.
Sources: airjelly.ai (product, features, and privacy claims); MakerStack's AirJelly review, 2026 (makerstack.co/reviews/airjelly-review); Kingy AI's hands-on Air Jelly review (kingy.ai); Sir John Whitmore, Coaching for Performance, 5th edition (Nicholas Brealey, 2017).
Related: How to Find Your Passion · Best Self-Improvement Books · How to Make Better Decisions · AI Coach App — Building It in 8 Hours