James Anderson

@JamesAnder007

Building in tech ⚡ | AI, cybersecurity & software | Sharing what I learn, build & discover | Always learning, always shipping

Dublin , 🇮🇪

Joined January 2026

Generative AI
James Anderson
@JamesAnder007
22h
Hot take: today's AI tools are genuinely useful AND genuinely a toy — at the same time. The reason it still feels like play: the intelligence is trapped in a chat box while all your real work happens somewhere it can't go. Full story here dev.to/james_anderson…
Startup Community
James Anderson
@JamesAnder007
22h
I counted my tabs mid-task the other day. 14 open. My AI was living in exactly one of them. It knew everything. It could touch nothing. I'd become the courier — carrying its answers to apps it can't reach. The full story here 👇
Are AI Tools Actually Making Us Productive — or Just Giving Us Something New to Play With?
From dev.to
James Anderson
@JamesAnder007
22h
I just published Are AI Tools Actually Making Us Productive — or Just Giving Us Something New to Play With? medium.com/@infoinletdev.…
Generative AI
James Anderson
@JamesAnder007
Aug 25
Before you swap in a bigger model to fix your RAG, check these: • chunks that split ideas in half • vector-only search (dies on exact matches) • no reranking (best chunk buried at #8) • no query rewriting • no retrieval eval The whole story here 👇
The Retrieval Checklist I Wish I'd Had Before Shipping RAG
From dev.to
Startup Community
James Anderson
@JamesAnder007
Aug 25
My RAG gave a confidently wrong answer, so I blamed the model. Swapped a bigger one. Tuned the prompt. No better. The model was fine. It was summarizing the wrong chunks. When RAG fails, it's retrieval ~73% of the time — not the model. read the full story: dev.to/james_anderson…