@romiteld

Firecrawl Part 2: The Confidence Gate That Decides When Bing Gets a Vote

2d ago · 9 min read · I found the bug in the most boring place: a company record that looked "complete enough" to pass a quick glance, but still had an Unknown-ish name and no usable location. The pipeline hadn't crashed. It hadn't even thrown a warning. It had just… stop...

Firecrawl Part 2: The Confidence Gate That Decides When Bing Gets a Vote

2d ago · 9 min read · I found the bug in the most boring place: a company record that looked "complete enough" to pass a quick glance, but still had an Unknown-ish name and no usable location. The pipeline hadn't crashed. It hadn't even thrown a warning. It had just… stop...

Tracing an Extraction Pipeline Like a Ledger: Trace Nodes, DLQ Boundaries, and Replayable Failures

3d ago · 11 min read · I noticed the same failure pattern twice: the extraction result was “wrong,” but nothing was actually broken. The model had simply taken a different branch. That’s the kind of bug that makes you mistrust everything—your cache, your prompts, even your...

Per‑Region PBR From One Photo: The Cropping Trick That Stops RGB‑X From Bleeding Materials Across Boundaries

4d ago · 10 min read · The roof was reflecting light like a painted wall. The input photo was fine. The segmentation was fine. The diffusion pipeline ran without errors. And yet the output didn’t pass the most basic human test: “does this look like the same material if I m...

My Three‑Phase Parallel Orchestrator: Typed Results, Exception‑Proof Phases, and a Rollout That Never Flaps

4d ago · 10 min read · I noticed it the first time I watched a voice session stall: the user had already finished speaking, but the system was still “thinking” like it was reading a novel line-by-line. That was the moment I stopped treating orchestration like a chain of ca...