Fetching from the wire…
Public story · 2026-07-25 · high
The stack also adds a runtime, a scheduler and a coding tool with a multi-agent mode, unveiled July 24 in Thailand.
Why now: Huawei announced the stack on July 24 at its Cloud Summit in Thailand, covered here a day later.
Huawei Cloud launched a four-part Agentic Infrastructure stack at its Cloud Summit in Thailand on July 24. The centerpiece is an Agentic Memory Storage Service that stores agent memory at petabyte scale.
That's built for agents running tasks over hours, not seconds, where losing context between steps breaks the job. Huawei sells that memory as its own service rather than something each application team has to build.
The other three pieces round out the stack. A UnifiedBus-based AI Cluster Service handles token generation, and AgentSphere is the runtime agents execute in. CCE VolcanoNext schedules general and AI compute together instead of running them as separate pools.
Alongside the infrastructure, CodeArts Agent moved into open beta, generating code at the project level instead of file by file. Its new Agent Team mode lets multiple agents work the same project at once.
The announcement doesn't say what petabyte-scale memory costs, or how latency holds up under concurrent access. That matters once several agents in Agent Team mode hit the same project's memory at once.
Each link below shares sources, entities, or timing with this story.
The Pragmatic Engineer published a deep read on August 25 of Inspect, the coding agent Ramp built instead of standardizing on Claude Code or Cursor. The numbers: Inspect authors 75% of Ramp's merged PRs, 90% of PRs in its own repository, passed 1 million total sessions in July...
A paddo.dev essay published August 23 argues the standard burnout research on AI coding describes the previous regime, since DORA has no 2026 report and there's no Stack Overflow 2026 survey, so the most-quoted studies describe a developer using an assistant rather than superv...
Hugging Face published its Summer 2026 State of Open Models report on August 14, and one statistic in it went almost entirely unremarked in the coverage. By July 2026, agents rather than humans became the Hub's primary users. Claude Code alone accounted for 44.4% of all agent...
The leaderboard says first place. The methodology says you should check your own bill. Qwen3.8 Max now ranks first on Artificial Analysis' agentic index, scoring 86.1 on OSWorld-Verified ahead of GPT-5.6 Sol Max at 83.2 and Fable 5 at 85.0, priced at $2.00/M input and $6.00/M...
DeepSeek-V4-Flash-0731 landed July 31 under MIT with a DSpark speculative-decoding module attached. Terminal Bench 2.1: 82.7. Toolathlon-Verified: 70.3. DSBench-FullStack: 68.7. DeepSWE: 54.4. NL2Repo: 54.2. The model card claims it beats DeepSeek-V4-Pro (Preview) "despite its...
GPT-5.6 Luna went to $0.20 input / $1.20 output per million tokens on July 30. That's an 80% cut. Terra dropped 20%. Luna's input now undercuts Gemini 3.1 Flash-Lite ($0.25/$1.50) and sits at one-fifth of Claude Haiku 4.5's $1 input. Simon Willison covered the announcement and...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.