ABØ Studios, a software/product studio
We take nearly-finished software across the finish line: launch rescue for AI-built web apps, and accessibility engineering for teams that must prove it. Inspect, fix, test, verify. Evidence, not promises.
An unfinished build is scattered layers. The scan finds the gap; the fix pulls the stack into one verified, shippable surface.
95.9% of top home pages failed WCAG (WebAIM Million, Feb 2026). 23/23 targeted tests passing (Cal.diy remediation). 317 tests (FrameCritic suite). READY WITH FIXES (LaunchCheck sample, 2026-09-12). LaunchCheck $49: one app, 24-hour target. No automated-only audit ever presented as a review
Two lanes, one standard
Founders and teams buy different things. The site never mixes them. Each door has its own offer, price, and next step.
Lane A: founders
Your app is 90% done. We find the launch blocker, prove it with evidence, fix it in a bounded scope, and re-verify it in a real browser.
Lane B: software teams
Fixed scope. Fixed price. Merged code. Evidence you can hand to a buyer. Real remediation in your codebase, with dated evidence a buyer can inspect.
The tension
The mechanism
Diagnosis earns the repair. Nothing is scoped cold, and no finding reaches a client without a human verifying it first.
Run the release in a real browser across mobile, tablet, and desktop. Record what is observable, not what is assumed.
Isolate the failing layer: markup, runtime, asset, integration, or deploy. One blocker, one cause.
Repair the named finding in a bounded scope. No rebuild, no silent expansion.
Re-run the failing flow in the same browser matrix. Console, page errors, and layout rechecked.
Capture before/after evidence a third party could follow. If it cannot be reproduced, it is not claimed fixed.
Hand over the fix order and evidence. Launch status stated plainly: ready, ready with fixes, or hold.
Selected proof
No clients to name yet, so the proof is the work itself: open tooling, a real remediation, and a dated sample with its limits stated.
OPEN-SOURCE TOOLING
Local-first TypeScript + Playwright + Express visual QA, v0.2.0, MIT. Multi-viewport runs, annotated screenshots, structured findings, HTML reports, console and page-error detection, overflow and broken-image checks. Internal leverage for every LaunchCheck. Inspected, never faked.
317 testsin the project's stated suite (verified against its README)
Inspect FrameCritic on GitHub ↗LANE B REMEDIATION
Real open-source scheduling app, real NVDA session. Date cells announced "11 button" before the fix and full date state after. Three components fixed, limitations documented, nothing claimed beyond what ran.
23/23 passingtargeted tests on the remediated components
Read the full teardown →Real sample — ABØ-owned app
Checked 2026-09-12 across mobile, tablet, and desktop in two browsers: two bounded findings (MEDIUM + LOW), no reproduced major blocker. $0. Not a paying customer, never presented as one.
READY WITH FIXESstated verdict, not a score
See what a $49 check covers →Lane B, in evidence
Cal.diy is a real, open-source scheduling app. A person using NVDA could not reliably tell which date was which. Here is exactly what that sounded like — before ABØ touched the code, and after.
11 button
A bare number. No indication it was a date, a button, available, or selected.
August 2026 grouping Tuesday, August 11, 2026, available toggle button pressed
Full accessible name, locale-aware, with selection state exposed via aria-pressed.
Read the full technical teardown →
95.9%
of top home pages had detected WCAG failures in February 2026 — up from 94.8%, the first regression in six years.
WebAIM Million ↗56.1
errors per page on average, up 10.1%, as page complexity rose 22.5% in one year.
WebAIM Million ↗25.55%
of Q1 2026 ADA web filings targeted sites that already had an accessibility widget installed.
EcomBack ↗How engagements run
Reproduce the blocker in a real browser. Name it, with evidence.
Bounded price, named surface, stated exclusions — agreed in writing before work.
Hands-on repair inside the agreed scope. No silent expansion.
Re-test the flow, capture before/after evidence a third party could follow.
Fix order, evidence pack, and a plain launch verdict: ready, ready with fixes, or hold.
The studio
The studio
Built by Abiola Obafemi, a 14-year-old software developer focused on turning ambitious ideas into working systems.
Small studio. Serious execution. AI-assisted, human-verified. Every automated signal is checked by a person before it reaches a client.
Launching soon? Start with the $49 diagnostic. Procuring software that must prove accessibility? Start with a pilot.