News, but built for builders.
Pillars over headlines.
Daily tech news has a structural problem: it pretends each headline is a self-contained event. It isn't. The interesting story is almost always a slow-moving constraint — supply chain, regulation, energy, talent — that surfaces as ten headlines spread across nine months, none of which make sense on their own.
I organize my reporting around those slow-moving constraints. Each one gets a pillar: a long-running, continuously-updated topic hub with an executive summary at the top, a tracker for the moving parts, a weekly-updated timeline, and twenty to thirty sub-articles drilling into specific angles. Sub-articles link back into the pillar; the pillar links out to the world.
The aggregated headline feed you used to come here for is still here — at the bottom of the homepage and the bottom of every pillar. It's titles-only, with one click to the original source. I don't paraphrase, I don't summarize, I don't intermediate the click. The point of that feed is to be useful, not to be the destination.
From news to pillar.
A pillar isn't born — it accretes. Here's the four-stage process every active pillar on this site has been through, and how we decide to open or close one.
Pattern surfaces
Three to five headlines, weeks apart, point at the same underlying constraint. I flag it and track it in private.
Pillar opens
An executive summary and tracker go live. I commit to a weekly update cadence. Sub-articles begin shipping.
Continuous edit
TL;DR is rewritten as the story evolves. Timeline updates Mondays. Solutions tracker refreshes when status changes — not on a schedule.
Pillar closes
When the underlying constraint resolves, I publish a postmortem and archive the pillar. It stays readable; it stops getting edits.
Read what came before.
Samsung's HBM moment, finally. And four other things from the bottleneck.
The CoWoS bottleneck the GPU shipment number won't tell you about.
The post-cloud cloud, two years in. Where the margin actually went.
Paged attention, two years later: the unreasonable effectiveness of one paper.
SMR contracts the hyperscalers can't finance. The first real test arrives.
Export controls vs. implementation reality. A line-by-line read.
The only tech publication I read end-to-end every week. The pillar approach should be the default, everywhere.— Senior PM, hyperscaler · Newsletter reader since 2023
It's the difference between drinking from the fire hose and looking at the river map. Both are useful; only one tells you what's happening.— Partner, infrastructure VC · Member since 2024
I cite the AI memory bottleneck pillar in every board deck I write. It's that good. The CoWoS tracker is irreplaceable.— Chief Strategy Officer, AI startup · Member since 2022
Built for the people actually building.
If that's you — engineer, founder, operator, investor — there's no better place to keep your map of the territory current.