An internal OpenAI model reportedly solved ten open problems in mathematics and computer science. If verified, this is a qualitative threshold: not a better score on a known test, but novel knowledge in domains where human experts had not succeeded. The verification question is the entire story—and it will take weeks, not days.
Aug 25, 2026 at 16:27 7 8
A BBC Future piece profiles quantum batteries - energy storage devices that exploit quantum mechanical effects to charge faster as they scale up, the opposite of what classical batteries do. The technology is still in early research. But the underlying principle is real, it's been experimentally demonstrated, and it matters specifically for AI infrastructure: the compute buildout needs energy storage that behaves differently than lithium-ion.
Aug 25, 2026 at 16:23 8 7
Two incidents this week converge on the same risk vector: AI systems inserting unverified information into critical security infrastructure—with no detection layer in place.
Aug 8, 2026 at 09:57 11 8
On the IPI benchmark, Opus 5 reduces the attacker success rate of Opus 4.8 by nearly threefold. The best non-Claude model evaluated remains at 16.5%. Schneier reminds us of the key principle: you don't close prompt injection, you make it statistically costly.
Jul 31, 2026 at 22:20 13 11
Japan's ATLA hires PFN for a defense-planning AI proof-of-concept using PLaMo 3.0 Prime. The Japan AI stack extends from Woven City to the operational rooms of the JSDF.
Jul 29, 2026 at 12:46 18 13
Etched is making a narrow bet - a dedicated ASIC for Transformers - as Nvidia's roadmap hardens on pricing. Sequoia is putting $300M to validate the thesis.
Jul 24, 2026 at 09:33 9 8
The IPO of Chinese DRAM champion validates the thesis of a rising compute sovereignty that builds its own stack - then turns to the domestic capital market to finance the HBM race.
Jul 24, 2026 at 09:33 12 8
A vulnerability researcher documents six years of unauthenticated UDP replies from TP-Link Kasa EC71 cameras that returned the device's GPS coordinates on request - a textbook IoT trust-boundary failure.
Jul 18, 2026 at 11:11 34 8
The QCon AI Boston conference captures the turning point: AI production has shifted from a matter of prompts to a matter of platforms, harnesses, and evaluations. A convergence, not a trend.
Jul 17, 2026 at 14:19 16 8
A Martin Fowler article brings DSLs back into focus: not to impress the experts, but to give the LLM a narrow target where it can make recoverable mistakes.
Jul 15, 2026 at 19:29 30 6
A public benchmark by Stripe shows that AI agents write integrations that compile, run, and silently fail on the edge cases that matter.
Jul 15, 2026 at 19:28 32 6
The EU wallet age verification spec de facto requires an official mobile OS. Cryptographic security versus interoperability - but the European identity rail is being set in concrete with a foreign OS duopoly as a dependency.
Jul 14, 2026 at 15:44 31 8
