most productivity tools are quietly hostile.

they pile up notifications, turn tasks into debts, and slowly stop feeling like help. at some point the thing meant to support you starts feeling like a manager reading over your shoulder.

nudge is one of the first things we built, and we didn't build it alone. it came out of working with el fin, a studio we co-create with. that's worth pausing on, because it's how the lab actually operates: capslock doesn't drop a finished tool on a company. we get inside how a company already works and design the system with them. nudge is what that looks like in practice.

it started as a boring question — how do we remind people about tasks — until we realized we were asking the wrong one. the teams we work inside weren't disorganized. they knew exactly what needed doing. what they were losing was momentum. their tools were great at tracking work and terrible at sustaining it.

so we threw the question out. not "how do we get people to finish tasks." "how do we help people feel like they're moving."

the flip

one decision changed everything: stop treating a notification as an alert, start treating it as a moment of reinforcement.

sounds like nothing. it's the whole thing. a reminder whose job is encouragement is a completely different object than one whose job is accountability. the language stops being about overdue work and starts being about progress. we stopped building something that pressures people. we started building something that keeps them going.

nobody needs more pressure. they need a reason to keep going.

how it actually works

small on purpose. it stays out of the way.

  • it runs on n8n, bolted alongside the tools a team already uses instead of replacing them.
  • it speaks through whatever channel the team already lives in — right now telegram and discord — so there's nothing new to adopt and no dashboard to forget about.
  • when something comes due, it nudges the one person who owns it, where they already are.

most of the work wasn't features. it was who, when, and tone.

  nudge ▾
 ╭───────────────────────────╮
 │ heads up — the deploy is  │░
 │ green. ship when ready.   │░
 ╰───────────────────────────╯░
  ░░░░░░░░░░░░░░░░░░░░░░░░░░░░░

                       09:24
fig. a nudge, in context

no magic

worth saying plainly: nudge isn't a big ai model running in the background, spending tokens to watch your team and decide things for them. the ai did its part earlier — it helped us build the thing — and then got out of the way.

what's actually running is unglamorous on purpose: a set of well-wired requests with good semantics and correct timing, shaped by behavioral science — triggers, small habits, a little game-like feedback. the intelligence lives in the system design, not in a model deciding for you.

we think most of the genuinely useful ai right now looks like this. less "the ai does it all for you," more "use ai to build a small system that does one real thing well." that's the posture of the whole lab.

and the economics make the point better than we can: the entire ai usage behind this costs about one cup of coffee a week, and el fin now produces roughly three times what it used to. the leverage was never in spending more on models — it's in wiring them up well.

what happened

  • the win wasn't the notification. it was killing the checking ritual — that background tab in everyone's head labelled "things to follow up on." the system handles it. you stop carrying the list in your head.
  • a nudge in a shared channel is gently public, so follow-through climbed on its own. nobody had to police anyone.
  • the first version pushed too hard. dialing in frequency and timing — so it lands like a good teammate and not an alarm clock — turned out to be the actual job. still tuning it.

underneath all of it sits the real lesson: friction is emotional before it's operational. people don't skip a task because no one reminded them. they skip it because the pile of unfinished work is heavy, and one more alert just makes it heavier. nudge pushes from the other side.

where it's at

honestly, nudge itself is still early — and we're not going to dress it up. the studio-wide gains are real, but we don't yet have hard metrics on nudge specifically. two signals we're wiring up to watch:

  • whether people actually engage with the nudges
  • whether overdue work goes down

the bet we're trying to prove: a little encouragement, timed well, beats endless reminders of obligation. every time.

small system. but it handed us a rule we keep building on — meet the work where it already happens, and design for momentum, not compliance. nobody needs more pressure. they need a reason to keep going.