Matthew O'Riordan

🧵 Microsoft is bringing AI to Premier League data. But the real challenge isn't the AI - it's the delivery. I built a demo to show what's actually possible today: football-commentary-ai.fly.dev Let me show you... (1/6)

I was hoping the @OpenAI Codex hype delivered. It really doesn't so far given it cannot run tests or use a pretty standard Node environment. Isn't the ability to actually run code inside the cloud environment table stakes for a programming agent? 🤯

Software development's biggest shift isn't AI writing code—it's the death of handoffs. When designers build in production and engineers design in code, the entire process collapses into continuous creation. The new skill: knowing when to ship & iterate. linkedin.com/posts/cl%C3%AD…

Loving Claude CLI @AnthropicAI, Cursor has been firmly relegated. But the persistent "Error during compaction: Conversation too long" is beyond frustrating because compaction IS needed because the conversation is too long 🙃

Had high hopes for @CuriousThingAI's Lucy, multi-lingual phone answering AI. Sadly it's a long way from a natural voice that sets up the caller for a better experience than a voicemail message. I eagerly await personal AI that means I no longer need to answer the phone

Shoutout to Ably for making real-time chat a DREAM! ☁️ Instant message delivery + presence tracking? Game changer! 🚀 Seamless communication across all devices is FINALLY here. Reliable & responsive chat = happy users.

Long conversations with @cursor_ai kill the IDE consistently. Small tasks are clearly better, but lengthy debugging cycles are a thing, and @cursor_ai needs to optimise for this given it translates directly into $$$.

Wow, Claude 4 Sonnet is mighty good at executing tasks, and Opus 4 exceptionally good at planning. What an impressive step change from the Claude 3.5 models from @AnthropicAI. These are my new favourite coding models. 🤓