SecondSource Morning Brief · September 4, 2026 | The gate's lifespan is set outside it, by ho…
1. Three frontier labs have moved their strongest security AI to an approved list. How long that gate stays shut is set by how many months the models
[Delayed: system issue] SecondSource Morning Brief · September 1, 2026 | That 14x measures routing, not the market
This issue arrived about 7 hours later than usual today — apologies for the delay.
SecondSource Morning Brief · August 31, 2026 | Money follows narrow tasks, not capability
1. Open-weight models — the kind whose parameters are published, so you can download and run them yourself — took 29% of the traffic at one
[Delayed: system issue] SecondSource Morning Brief · August 30, 2026 | Marvell's filing: warrants cut its own reven…
This issue arrived about 9 hours later than usual today — apologies for the delay.
SecondSource Morning Brief · August 29, 2026 | Watch list price and rate limits, not revenue
1. Revenue growth cannot tell you whether an AI lab has pulled compute back from the open market; the number looks the same under two opposite
[Delayed: system issue] SecondSource Morning Brief · August 27, 2026 | An unsolvable eval problem is where an AI st…
This issue arrived about 3 hours later than usual today — apologies for the delay.
SecondSource Morning Brief · August 26, 2026 | Acceptance-test models in your own environme…
1. One model, one change of evaluation setup, 91 → 99: model acceptance testing has to happen inside your own environment.
[Resend: delivery issue] SecondSource Morning Brief · August 25, 2026 (sent late) | Check the accounting basis before you compare
This issue arrived about 7 hours later than usual today — apologies for the delay.
SecondSource Morning Brief · August 24, 2026 | Ask: does the cheapest config offload KV?
1. For the first time, the gap between NVIDIA and AMD can be measured in public, by anyone willing to re-run the numbers. The dividing line is not
SecondSource Morning Brief · August 23, 2026 | "AI opposes regulation" — united, minus one
This issue arrived about 3 hours later than usual today — apologies for the delay.
[Resend: delivery issue] SecondSource Morning Brief · August 22, 2026 | Which compute layer is that price from?
This issue arrived about 4 hours later than usual today — apologies for the delay.
SecondSource Morning Brief · August 21, 2026 | Price the warrant as a discount on orders
1. We got our own lead story backwards yesterday. The warrant Google holds is a discount given by the seller, and it runs 5 to 6.5 percent of what
SecondSource Morning Brief · August 20, 2026 | Ask your chip vendor for an at-market warrant
This issue arrived about 2 hours later than usual today — apologies for the delay.
No briefing today (Aug 19) — our apologies
Hi,
SecondSource Morning Brief · August 18, 2026 | Optics timing goes from rolling window to na…
AMD moved "copper in the rack becomes optical" off a three-year window that never expires and pinned it to a named 2027 product. The supply chain
SecondSource Morning Brief · August 17, 2026 | Disclosure clustered, not incidents — counts…
Every incident in August's wave of "AI models reaching real companies" actually happened between April and late July — so ranking labs by incident
SecondSource Morning Brief · August 16, 2026 | Memory, not wafers, drives AI hardware cost
TSMC conceded the supply gap is "very big" and still refused a multiples-scale price rise — and the money it declined went to the memory makers.
SecondSource Morning Brief · August 15, 2026 | The capex-to-chip ratio is rising, not falli…
The share of the big four cloud companies' money that becomes AI chip revenue is rising, not falling: 24.5%, 28.5%, 29.2% — the opposite of where we
SecondSource Morning Brief · August 14, 2026 | Speed gains cap at the model's share of time
OpenAI's Ultrafast tier claims 14x faster — measured against the company's own standard tier. Academic profiling puts tool processing at 50% to 90%
SecondSource Morning Brief · August 13, 2026 | Tomorrow a classifier approves, not you
From tomorrow, Claude Code stops asking a human to sign off on each action by default. A classifier reviews instead. Anthropic's stated reason is