Over the last 10 days, I competed in the Global AI Developer Hackathon hosted by @Meta and @cerebral_valley. I produced Parity - an agentic workflow which closes the gap between what a sighted consumer sees, and what a blind consumer hears. Check out my demo!
Built in Cape Town πΏπ¦ for Meta's global AI hackathon: an agent that finds where online checkouts fail blind shoppers, before customers or lawsuits do. 70% of 2025's US accessibility suits targeted e-commerce. Would love your support π https://t.co/R5HMwUMqgW
Over the last 10 days, I competed in the Global AI Developer Hackathon hosted by @Meta and @cerebral_valley. I produced Parity - an agentic workflow which closes the gap between what a sighted consumer sees, and what a blind consumer hears. Check out my demo!
Your checkout can pass axe-core with 0 violations and still be unusable with a screen reader. Parity shows the gap side by side, then proposes a minimal HTML fix and re-checks it. Try it, no login: https://t.co/CXLQJcNggN https://t.co/R5HMwUMqgW
Over the last 10 days, I competed in the Global AI Developer Hackathon hosted by @Meta and @cerebral_valley. I produced Parity - an agentic workflow which closes the gap between what a sighted consumer sees, and what a blind consumer hears. Check out my demo!
Small but honest eval: 38 seeded accessibility faults that axe-core passes, plus 20 clean controls. Full system 85% recall / 99% precision / 0 false positives. Ablations and a per-stage sweep across 7 models in the repo. https://t.co/R5HMwUMqgW
Over the last 10 days, I competed in the Global AI Developer Hackathon hosted by @Meta and @cerebral_valley. I produced Parity - an agentic workflow which closes the gap between what a sighted consumer sees, and what a blind consumer hears. Check out my demo!
"The model perceives; code decides." One agent, gated tools, six code checks, a judge, a fixer and a sandboxed re-audit. Ablations: screenshot only 17% recall, tree only 75%, both plus the code gate 85% at 99% precision. https://t.co/R5HMwULSro
Over the last 10 days, I competed in the Global AI Developer Hackathon hosted by @Meta and @cerebral_valley. I produced Parity - an agentic workflow which closes the gap between what a sighted consumer sees, and what a blind consumer hears. Check out my demo!
Computer-use agent with code as the referee: Muse Spark walks the checkout, a grounded model compares screen vs accessibility tree, and six deterministic checks reject any claim it can't prove. A human approves every fix. π https://t.co/R5HMwUMqgW
Over the last 10 days, I competed in the Global AI Developer Hackathon hosted by @Meta and @cerebral_valley. I produced Parity - an agentic workflow which closes the gap between what a sighted consumer sees, and what a blind consumer hears. Check out my demo!
Screen reader users: I built a tool that compares what's on screen with what VoiceOver actually announces (e.g. a "Pay R1,249" button read as just "Submit"). I'd really value your view on whether these are the gaps that matter most to you. https://t.co/R5HMwUMqgW
Over the last 10 days, I competed in the Global AI Developer Hackathon hosted by @Meta and @cerebral_valley. I produced Parity - an agentic workflow which closes the gap between what a sighted consumer sees, and what a blind consumer hears. Check out my demo!
Automated checkers catch a fraction of real barriers. On my test checkout axe-core reports 0 violations, yet VoiceOver announces the "Pay R1,249" button as just "Submit". Parity compares what a sighted user sees with what a screen reader hears, proves each gap, and proposes a fix a human approves. Feedback very welcome π https://t.co/rHkxvrfYZ9
Over the last 10 days, I competed in the Global AI Developer Hackathon hosted by @Meta and @cerebral_valley. I produced Parity - an agentic workflow which closes the gap between what a sighted consumer sees, and what a blind consumer hears. Check out my demo!