Builds

EXP-009Building
Mullo
The best page, the worse journey: when an AI tool's quality isn't worth the effort
An AI that draws my notes as an illustrated page. Against a general-purpose tool it made the better page but took far more effort, so I'm redesigning it around what only it does.
Read the log
EXP-007Building · Pilot
This lab, rebuilt
Nothing goes live unread: building a public site with AI agents, without losing control
This lab, rebuilt in three days by Claude Code from briefs I worked out with Claude. The setup that keeps an agent-built site honest, and the rules that came from mistakes.
Read the log
EXP-006Building · Pilot
Fold's review panel
Never mark your own homework: independent AI reviewers for a solo builder
Independent AI reviewers for an app I was building faster than I could read its code. They caught bugs that had passed the tests, and one finding changed what Fold is.
Read the log
EXP-005Building · Pilot
Fold
Cheap to build, hard to stop: when AI builds faster than you can learn
A recipe app for vegetarian households with a vegan at the table, built for my own family's needs; once I saw its wider potential, I added a fair amount of complexity without checking whether it was warranted.
Request an invite Read the log
EXP-004Live
emilysas.com
The bottleneck moved: where the time goes when an AI agent writes the code
A Claude agent wrote my professional site in about 42 hours, from first commit to the last launch change. The time went on accounts, decisions only I could make, and testing in production.
Try it (opens in a new tab)Read the log

Earlier builds

EXP-001 · emilybuilds on Notion + Super
Deciding not to build: when no-code is the right call, and when it stops being one
In 2025 I published emilybuilds.ai on Notion and Super.so to start writing sooner; it was live within days, and the complexity moved somewhere I had less control over.
RetiredRead the log
EXP-002 · Launchling
All tests passing: when checking an AI product outgrows the product
An AI tool that turned a founder's rough idea into a plan. The product ran end to end; the evaluation I built to check it grew until it couldn't finish a run.
RetiredRead the log
EXP-000 · Ambient care-plan scribe
Before care had scribes: drafting care plans with AI in 2022
An AI that drafts a care plan while the assessment is still happening, begun in late 2022. Getting it to write was easy; knowing how far to trust the draft wasn't.
ParkedRead the log

What I build with

The tools behind everything in the lab, grouped by the job they do, with what I used in 2025 beside what I use now. Dates are to the month.

Compare my stack now with:2025: Launchling and the first site

switch moved to a different tool
stick same tool, same use
shift same tool, used differently

Thinking and planning
2025: Launchling and the first site
- Planning my 2025 builds, from architecture to prompt design. Claude is now my default; I still use ChatGPT to pull out context from previous tasks, and to compare outputs: I send the same prompt to ChatGPT and Claude, and sometimes pass each one's output to the other to combine the best of both.since Dec 2022
What changed
From a chat that wrote the instructions I pasted into the editor, to a planning session that writes specs into the repository for the builder.
My stack now
- Planning my 2025 builds, from architecture to prompt design. Claude is now my default; I still use ChatGPT to pull out context from previous tasks, and to compare outputs: I send the same prompt to ChatGPT and Claude, and sometimes pass each one's output to the other to combine the best of both.since Dec 2022
- Planning, briefs and specs for each build, and editing my writing.since Jan 2026
Writing the code
2025: Launchling and the first site
- The editor I built Launchling in; its AI wrote most of the code, from the front end to the tests.Jul 2025 to Aug 2025
What changed
From an editor whose AI wrote most of the code but which I had to stop from running scripts, to an agent trusted to build, run and test in its own session.
My stack now
- The agent that writes, runs and tests the code for Fold, Mullo and this lab.since Jun 2026
Design
2025: Launchling and the first site
- Design in 2025. For emilysas.com in October 2026 I used GPT 6 for brand and logo design only, as a second set of directions to critique against Claude's. Claude is now my default; I still use ChatGPT for image generation.since Dec 2022
What changed
From layouts I had to fix for phones after launch, to a design tool and layout tests at phone and iPad sizes.
My stack now
- Design in 2025. For emilysas.com in October 2026 I used GPT 6 for brand and logo design only, as a second set of directions to critique against Claude's. Claude is now my default; I still use ChatGPT for image generation.since Dec 2022
- Screenshots and layout checks at phone and iPad sizes, run when a change touches the app.since Oct 2026
- Designing Mullo's screens before building them.since Oct 2026
Checking the work
2025: Launchling and the first site
- Launchling's system for scoring AI-written plans.Jul 2025 to Aug 2025
What changed
From one evaluation system I couldn't keep running, to small reviewers that each check one thing.
My stack now
- Independent reviewers for security, privacy, QA, cost, analytics and recipe truth (how).since Oct 2026
Research
2025: Launchling and the first site
- Research for my 2025 builds. Claude is now my default.Dec 2022 to Aug 2025
What changed
From asking the same chat to a separate research tool.
My stack now
- Desk research with sources: a little from March 2026, then heavily from late October 2026, much of it for Fold.since Mar 2026
Models inside my products
2025: Launchling and the first site
- Writing Launchling's plans and judging them.Jun 2025 to Aug 2025
What changed
From one model for everything to a different model for each job.
My stack now
- Opus reads recipes in Fold and Haiku checks its drawings; Sonnet 5.5 writes Mullo's plans and hints.since Oct 2026
- Nano Banana 2.1 draws Fold's dishes and Mullo's pages (Pro for comparison), and Gemini 3.8 Flash checks Mullo's pages.since Oct 2026
Back end and automation
2025: Launchling and the first site
- Launchling's back end, and Fold's today.since Jul 2025
- Connecting forms, the model and email before Launchling had a back end, and the old site's sign-ups.Jun 2025 to Nov 2026
What changed
From wiring services together in Zapier to a back end written in code.
My stack now
- Launchling's back end, and Fold's today.since Jul 2025
Email and sign-ups
2025: Launchling and the first site
- Email on my own domain.since Jul 2025
- Sending Launchling's emails at first.Jul 2025
- Sign-up and onboarding forms.Jun 2025 to Nov 2026
What changed
Sign-ups go straight into the newsletter tool, with no automation in between.
My stack now
- The newsletter, with the sign-up form built into emilysas.com.since Oct 2026
- Email on my own domain.since Jul 2025
Websites
2025: Launchling and the first site
- Writing the pages of the old emilybuilds.ai, published with Super.so.since Jul 2025
- Turned the Notion pages of the old emilybuilds.ai into a website, with analytics, for $26 (about £19) a month.Jul 2025 to Nov 2026
What changed
From a site I had less control over to one I own, edited in a content manager.
My stack now
- The framework for emilysas.com and this lab.since Oct 2026
- Where I edit both sites' content, with every change saved to Git.since Oct 2026
- Writing the pages of the old emilybuilds.ai, published with Super.so.since Jul 2025
App front end
2025: Launchling and the first site
- Launchling's first version, until its limits and pricing pushed me to code.Jun 2025 to Jul 2025
- Launchling, then Fold.since Jul 2025
What changed
Launchling moved from Framer to code in 2025; Fold has been code from the start.
My stack now
- Launchling, then Fold.since Jul 2025
Hosting
2025: Launchling and the first site
- Hosting for Launchling, then Fold.since Jul 2025
What changed
Sites moved to Netlify; apps stayed on Vercel.
My stack now
- Hosting for emilysas.com and this lab.since Oct 2026
- Hosting for Launchling, then Fold.since Jul 2025
Data
2025: Launchling and the first site
- Launchling's plans, feedback and evaluation results, and the old site's sign-ups.Jun 2025 to Nov 2026
What changed
From a spreadsheet-style database to hosted databases and private repositories.
My stack now
- Fold's household data.since Oct 2026
Analytics
2025: Launchling and the first site
- Visits to the old site.since Jul 2025
What changed
Analytics that need no cookie banner.
My stack now
- Visits to the old site.since Jul 2025
- Visits to emilysas.com, without cookies.since Oct 2026
Domains
2025: Launchling and the first site
- My domains, including emilybuilds.ai and emilysas.com.since Jul 2025
What changed
Unchanged.
My stack now
- My domains, including emilybuilds.ai and emilysas.com.since Jul 2025
Code and CI
2025: Launchling and the first site
- Checks on code changes (Launchling's evaluation tests; Fold's tests).since Jul 2025
- Code, and some of Fold's data, in private repositories.since Dec 2014
What changed
The same tools, with far more running through them.
My stack now
- Checks on code changes (Launchling's evaluation tests; Fold's tests).since Jul 2025
- Code, and some of Fold's data, in private repositories.since Dec 2014

Builds

Earlier builds

What I build with

Thinking and planning

Writing the code

Design

Checking the work

Research

Models inside my products

Back end and automation

Email and sign-ups

Websites

App front end

Hosting

Data

Analytics

Domains

Code and CI

Dispatches from the lab