InAgent achieves 90.2% on OSWorld: the computer-use agent gap narrows for the Chinese stack
A Chinese CUA becomes the first to surpass 90% on OSWorld. The gap with US frontier labs is now a harness gap, not a model gap.
4 h ago 5 6
A Chinese CUA becomes the first to surpass 90% on OSWorld. The gap with US frontier labs is now a harness gap, not a model gap.
4 h ago 5 6
DeepSeek pushes V4-Flash-0731 to public beta with Responses API, Codex adaptation, 82.7 on Terminal Bench 2.1 and 54.4 on DeepSWE. The vendor lock runtime decreases by one level.
Jul 31, 2026 at 12:50 12 10
A field guide argues embedded SQLite can carry a real production workload - if you know the WAL/mmap/VFS knobs. Under the hood, plus the trade-offs it won't fix.
Jul 29, 2026 at 12:46 11 13
Samsung plans to build the base die of HBM5 using its 2nm process, aiming for a 50% speed improvement. Memory is no longer just about increasing capacity; it's also about leveraging advanced logic-node economics.
Jul 27, 2026 at 10:16 14 12
Anthropic officializes Claude Opus 5, the culmination of the "Fable 5" cycle - announced capacity nearly equivalent for a price halved. The battle shifts to pricing.
Jul 26, 2026 at 18:36 6 8
A project posted on ziggit.dev announces a drop-in replacement for Bun, rewritten in modern Zig, with sub-1s incremental builds. The question: does the JavaScript runtime really need a fork?
Jul 25, 2026 at 10:27 10 9
Five days after making Fable 5 permanent in premium plans, Anthropic ships Opus 5 - a model close to flagship capabilities at about half the price per Tech in Asia. A pricing move that betrays competitive pressure.
Jul 25, 2026 at 10:27 10 11
ETNews details Xiaomi's humanoid running three concrete production tasks - the embodied race gets a China-side number breakdown.
Jul 19, 2026 at 16:34 9 11
Sugon officially unveils Dawn 8000 specs at WAIC: 20x compute density per unit, full-precision interconnect for 100,000 cards. The Chinese compute stack meets industrial standards.
Jul 17, 2026 at 14:19 19 8
Google differs Gemini 3.5 Pro to improve coding. The metric that blocks a frontier release is no longer knowledge, it's tool-use.
Jul 17, 2026 at 09:19 25 11