Damian Pickett

The problem

SmartSurvey already ran consultations for councils across the UK, but a capability isn't a product. Citizen engagement needed a name, a shape and a market.

The decision

Start in the market, not the feature list. I researched how councils run consultations, built the ICPs and personas for the officers answerable to residents, members and auditors at once, and positioned SmartCE as the way to listen to your community.

What I cut

The obvious category, consultation management software. Accurate, and forgettable: it sounds like paperwork.

The result

Every demo, sales tool and web page inherited the position, and the go-to-market campaign and direct outreach went to the exact profiles the research described. DigiGov put the story in front of public-sector buyers in person.

Read the launch story → · SmartCE ↗

The problem

AutoCategorise uses AI to sort open-text survey responses into themes. The capability was complex, and a launch had to make it simple to understand.

The decision

Let people see it work. With product and design I launched it with a product walkthrough, an interactive demo and an ROI calculator, so a buyer could watch it in action and put a number on its value.

AutoCategorise ↗

The problem

What customers and the market say lands with customer success and marketing first. It only shapes the product if someone carries it across.

The decision

Make carrying it across a habit. I run product workshops with customer success, the pod lead and engineers, and a feedback loop that brings customer and market input into product decisions.

The problem

Habit trackers turn wellbeing into admin: numbers, streaks and guilt. I wanted the Tamagotchi feeling instead, a creature you glance at and understand.

The decision

Planty reads the signals the phone already trusts, HealthKit and Screen Time, through plain rules you can read. A walk, a good night's sleep or an hour of focus helps him; too long in the apps you chose to watch makes him thirsty.

What I cut

Accounts, the cloud and AI guessing at your life, so everything stays on the phone. Streaks and death went too. Planty never dies, and one bad evening can't undo weeks of growth.

What changed

After the first TestFlight build in September I caught myself opening it less: Planty was lovely to look at and gave me too little reason to come back. So I reworked the core loop before polishing a single screen. He now has a song for you each day and a seed game to play, and the first version of that game went in the bin for being boring.

More on Projects → · A Something Wonderful Studios ↗ project

The problem

Soar scores my training week against targets and reads Apple Health, so walks, Zone 2, intervals and padel count without a tap. Lifts were the exception: running one meant picking up my phone between every set.

The decision

The Watch records the lift itself, one set at a time, and the phone stays the single store. Each set is saved on the Watch and sent straight across, so a phone left in a locker loses nothing, and the phone pairs the lift with its Health workout so nothing counts twice.

What I cut

An App Store release, accounts and analytics, because it's built for my own training. Per-set effort ratings, mesocycles and deloads are parked for later.

What changed

When I started a new training programme in October, I reworked Train around it rather than rebuilding it. It now opens on the day's session, with the generated workouts kept as a fallback.

The problem

Reminders lives on Apple devices. I needed the same list on my phone, in Chrome at work and within reach of AI assistants. So I built one, around a single idea: a task has four states, drawn as one box that fills as the work does.

The decision

One small server, one token, three clients. A Node app with no dependencies runs on Vercel with a Turso database, the iPhone app works offline and syncs, and an MCP server lets Claude and ChatGPT add tasks by name. I ruled out CloudKit, because a work browser without an Apple ID and an assistant both need plain HTTP.

What I cut

Repeating tasks and notifications stayed out of the first version until Progress earned them. Assistants can add tasks but never create lists, so the structure stays mine.

What changed

The morning after it went live I missed repeating tasks and drag to reorder, so both went in before lunch, and a Lock Screen widget followed that afternoon. The four-state box stayed exactly as it was.

The problem

News apps decide what you see with feeds and paywalls. I wanted one finite read each morning, made only of what I care about.

The decision

A Claude editor reads 55 sources I chose and picks the day's stories under written rules: two sentences each, then a link to the outlet that reported it. It works only from that morning's feeds. If the feeds fail, it publishes nothing and tells me why.

What I cut

An app, accounts and dependencies. It's a static site made of a fetch script, a build script and one file of editorial rules.

What changed

The first front page ran far too long, so it now opens on the highlights with the whole edition a tap away. One podcast pick a day became three.

Read today's edition ↗ · Source on GitHub ↗