SaaS engineering, AI development, and the unglamorous lessons from shipping software solo in Europe — documented in public, as it happens.
Content generation SaaS. One topic → blog post, LinkedIn draft, Twitter thread, email sequence, video script — in one pass. Built around a multi-format prompt pipeline with per-site AI key routing and automatic fallback across 14 OpenAI accounts + Anthropic + Groq.
Automated short-form video pipeline. Topic → AI script → TTS voice → rendered 9:16 video → YouTube upload. FFmpeg compositing driven by a BullMQ queue. Handles concurrent jobs per workspace with isolated OAuth tokens.
Affiliate link management SaaS. Cloaking, click tracking, geo-redirect, split-testing, and a dashboard that pulls real conversion data. Dead simple tool that does one job without adding a CRM nobody asked for.
Discussion tracker across Reddit, HN, forums, and RSS. Crawls sources, groups threads into semantic clusters using TF-IDF hashing — no embedding API. Groq for summaries. Ships as open source.
Google NotebookLM is genuinely impressive. Drop in a pile of PDFs, research papers, or meeting transcripts and it produces grounded summaries, answers with citations, and — its most-talked-about trick — a sur
Tonight I attended an AI, Agents & NASA event at Trinity College Dublin. The talks focused on agent-based AI systems, scientific research workflows, and how organizations like NASA are exploring AI to accelerate know
Lately I’ve been spending more time reading Reddit during outages, breaking product updates, infrastructure incidents, and AI discussions. One thing I keep noticing: The same conversation fragments into dozens of s
I run 14 editorial AI workspaces. Here's how I added a self-hosted LLM proxy to cut paid API calls with no changes to my app code.
No hype, no funnels. Engineering notes and systems for developers who want to ship with Claude — not just watch demos.