Home / Articles / Native apps by agents: four iPhone and iPad apps from one workspace and an Android tablet app in five days, from 1,164 lines of Swift to 48,230, a signed Android build four hours after the first commit, and a guide to the fringe on TestFlight ten and a half hours after its brief
Native apps by agents: four iPhone and iPad apps from one workspace and an Android tablet app in five days, from 1,164 lines of Swift to 48,230, a signed Android build four hours after the first commit, and a guide to the fringe on TestFlight ten and a half hours after its brief
By Dinis Cruz · 2026-11-15 · article v1.0.0 · newsroom v0.14.103 · iosandroidappsagentstestflightweb-summitpromptugalseriesarticle
Abstract: Around Web Summit, our agents built native apps, which I had not expected them to do so fast. On 7 November a small iPhone app, 1,164 lines of Swift, was imported from a vault into its own repository; by 11 November that workspace held four apps sharing thirteen Swift packages and 48,230 lines of code: the public PrompTugal reader, the team's own app, a RiskMandate app, and a guide to the Lisbon Tech Fringe. Every push to the development branch uploaded a build to TestFlight. Apple rejected one upload for a missing purpose string, and a good build followed an hour and twenty-five minutes later. On 8 November at 20:33 an Android app for the RiskMandate stand was started from nothing, and four hours and eleven minutes later a signed build was installed on my tablet; by the end it had 33,042 lines of Kotlin, an on-device Gemma chat and a stand mode. On 11 November the fringe guide went from brief to a valid TestFlight build in about ten and a half hours. This is the fourth article in the Ten days series.
In short
- Four iOS apps from one workspace, from 1,164 lines of Swift on 7 November to 48,230 by 11 November.
- TestFlight on every push to the development branch, and a first build approved for testers in about a day.
- An Android tablet app from nothing: a signed build on my tablet four hours and eleven minutes after the first commit.
- A fringe guide in a day: from brief to a valid TestFlight build in about ten and a half hours.
- Agents in many sessions, on separate branches, with conductors merging green branches.
- Part of a series: this is article four of Ten days at Web Summit.
From one small app to four
At 01:41 UTC on 7 November, a small iPhone app, version 0.3.1 and 1,164 lines of Swift, was imported from an sgit vault into a new repository. It was the start of PrompTugal's iOS workspace. In the first two days the agents made 517 and then 623 commits, and by 11 November the workspace built four apps from shared code:
- PrompTugal, the public reader app: the news, the agenda in Lisbon time, your own day, reminders, offline reading, and a private vault opened with Face ID.
- Team, the team's own app: dashboards of people, agents and decisions, briefings, and messages through encrypted vaults.
- RiskMandate, the company's app.
- Lisbon Fringe, the guide to the summit's fringe events.
By then the workspace had 13 shared Swift packages and 48,230 lines of Swift, all written by agents, and about 1,470 commits across its branches.
TestFlight on every push
The release pipeline came first. By 22:47 on 7 November, every push to the development branch built the apps and uploaded them to TestFlight, Apple's channel for test builds. The first build was approved by Beta App Review in about 24 hours. Apple also rejected an upload on 8 November, for a missing purpose string explaining why the app uses speech recognition; a good build followed an hour and twenty-five minutes later. At 23:20 that night the public app was submitted to the App Store.
The builds needed Macs, and that turned out to be the real constraint: the agents spent far longer waiting for a Mac than building on one. That is the next article in this series.
Many sessions, separate branches
The work was spread across sessions, each on its own branch. The build log for one night describes seven sessions, five building the app, one releasing it and one editing the site, with a conductor merging the branches that passed into the development branch about once an hour. There were dozens of role branches, and commits named the agents that made them: conductors, a release agent, and agents from Indi Gupte's and Silvio Switha's teams as well as mine. Indi's agents later submitted more builds for testers on their own.
An Android app for the stand
At 20:33 on 8 November, the Android app for the RiskMandate stand did not exist. At 00:44 on 9 November, four hours and eleven minutes later, a signed build was installed on my tablet, and the next one arrived as a one-tap update through Firebase. The next day was its busiest: 935 commits.
It was built for a tablet on the stand, in Kotlin, and it grew to 33,042 lines. It has a stand mode that pins the screen and resets when idle; a step-through demo of Agent Behaviour Policies in every layer; and a chat with Gemma, a small open model, running on the tablet itself, at about five tokens a second in the one run that was measured. It went to the team's tablets through Firebase App Distribution, not through Google Play, which I decided to leave for later.
A guide to the fringe in a day
On 11 November, the day after Storm Charles made Web Summit call off its second day, the brief for a Lisbon Fringe app arrived at 10:45. Seven minutes later there was code; in the evening, Indi's design replaced the first version; and at 21:11 the Fringe app was valid in internal TestFlight, about ten and a half hours after the brief. It carried 173 events and a filter for the events the storm had moved, from the same data as the fringe guide's website.
What it taught us
- Native apps are within reach of agents: four iOS apps and an Android app in five days, from a small start and from nothing.
- Ship the pipeline first: TestFlight on every push meant every change could be tried on a phone the same day.
- Shared code pays: four apps from one workspace and thirteen packages.
- The bottleneck moves to the machines and the stores: Macs to build on, and review queues to wait in.
What this does not claim
- Figures are from the repositories on 15 November 2026; line counts are raw lines of Swift and Kotlin files.
- Commit counts are approximate across branches; most early commits carry no agent name, so per-agent counts are partial.
- The Gemma speed is from one measured run.
Where this comes from
The repositories PrompTugal__App__iOS, PrompTugal__App__Android and PrompTugal__App__Mobile, read 15 November 2026: their commits, briefs, status files and workflows. Pages: ios.app.promptugal.news, its three apps and build log, and VoiceDebrief. Earlier on this site: ten days at Web Summit, the team's operating system, PrompTugal.news, the newsroom we ran.
Peer review Not reviewed yet. Review it
Threads
Builds on
- Ten days at Web Summit: what six people and around 250 agents built between 3 and 13 November, in order, and the series of articles that will tell each part Six people and around 250 agents, ten days at Web Summit: what we built, in order, and the series that will tell it.
- The team's operating system: web-summit.sgit.ai, the one site where six people and their agents ran Web Summit, with a brief and a command centre for each person, a key and an inbox for each agent, the board as files, and the outside world on a map web-summit.sgit.ai: one site where six people and their agents ran Web Summit, in four layers built on git.
- PrompTugal.news, the newsroom we ran at Web Summit: desks of agents that captured, mapped, wrote and reviewed 73 signed articles in ten days, an accountant that counted every run, and fifty outreach drafts that no agent was allowed to send PrompTugal.news: 73 signed articles in ten days from desks of agents, a reviewer for every draft, and fifty outreach drafts no agent could send.
Continued by
- Ten days at Web Summit: what six people and around 250 agents built between 3 and 13 November, in order, and the series of articles that will tell each part Six people and around 250 agents, ten days at Web Summit: what we built, in order, and the series that will tell it.
Article No. 417 · start a reading list with it
Reviews
No reviews yet.