The Stack - SiliconNoon (Page 6)

The Stack

Open Source Radar — September 17: voice, audits, and MCP testing

Today's trending board is mostly agents and the skill packs that feed them, so the interesting movement is one layer down — the tools that give an agent a voice, test the servers it talks to, and check whether the code it just wrote is actually safe. Voicebox (TypeScript, ~54,700 stars, MIT) — A local-first voice studio that covers both halves of the voice loop: cloning and speech generation on the way out, dictation into any text field on the way in, which is why it reads as an open answer to

The Stack

GLM's Infra Agent built the stack that serves GLM

Z.ai published the engineering log behind GLM-5.3-Flash today, and the news is not the cluster — it is who did the systems work. The lab says a GLM-5.3-powered "Infra Agent" wrote much of the inference stack that now serves that model, and it frames the result as an early, limited form of recursive self-improvement. Z.ai says all production inference for GLM-5.3-Flash runs on a cluster of more than 100,000 Chinese-made accelerators, and that much of the work of getting it there was done by a GL

The Stack

Nadella's lock-in test: pull out a model, see if your evals survive

Microsoft's chief executive spent a long interview arguing the AI fight is not about which model wins, and Apple published the factory-to-photo chain of trust behind its provenance system. Satya Nadella used a rare All-In podcast appearance to say the durable AI business is the harness, not the model — and that enterprises must own the weights, memory and evaluations their agents run on. "I want my privacy. I want to be able to embed my knowledge in a set of weights that I control. I want to se

The Stack

Open Source Radar — September 16: harnesses, skills, and voice

Today's board splits cleanly: the agent harness and the skill packs that fill it keep climbing, the prompts those agents run on are now public record, and two self-hosted platforms — one for knowledge, one for phone calls — showed up with real traction. Pi (TypeScript, ~106,000 stars, MIT) — The top of today's trending board is an agent harness rather than a model. Pi is a monorepo that ships four things separately: a unified multi-provider LLM API, an agent runtime with tool calling and state

The Stack

Workiva ships no-code agents for work that has to survive an audit

Enterprise AI's next fight is not capability, it is defensibility. Workiva spent its annual conference selling the version of agents that ends in a signature. Workiva put its agent layer in customers' hands at Amplify 2026, unveiling Agent Studio — a no-code capability for building, customizing and deploying AI agents inside its own reporting platform — alongside three agentic products for regulatory filings and a fourth for internal audit testing. The pitch is not a smarter model, it is a gove

The Stack

Cursor's 7.5x renewal ask is pushing customers to the harness layer

AI coding bills stopped being model bills this year. The enterprise contracts now moving tell you where the money is actually going. One IT consultancy paid Cursor about $200,000 for 800 licences in the year to May. At renewal, Cursor asked for roughly $1.5 million for the same usage — a 7.5x jump. The firm negotiated the ask down to $250,000, a quarter of the original price but still more than it paid last year. That number comes from The Information's reporting on Cursor's move to usage-based

The Stack

StackGen puts one governance plane under every production agent

Coding agents made change volume explode. Operations absorbed the blast without a comparable upgrade — and StackGen's answer is to stop bolting agents onto four silos and put one operating system underneath all of them. StackGen launched its Autonomous Operations Factory on Tuesday, a control plane that runs infrastructure, DevOps, observability and reliability agents against a single shared model of what is deployed, what changed, what broke and what fixed it. The foundation is Aiden OS, which

The Stack

Open Source Radar — September 15: agents get eyes and a voice

Today's board is about what sits between an agent and everything else: the hookup to the live internet, the voice it speaks with, and the box it runs in when there is no network at all. Agent-Reach (Python, 81,783 stars, MIT) — A capability layer that gives any command-running agent read and search access to the platforms where the useful information actually lives: X, Reddit, YouTube, GitHub, Bilibili, XiaoHongShu, RSS. Each platform gets an ordered list of backends — a primary plus fallbacks

The Stack

NVIDIA open-sources an agent harness that redesigned itself

NVIDIA's research arm published the code for a loop where agents studied other agents' transcripts and cut their own token bill. Meanwhile South Korea's central bank measured who actually gets the AI jobs. NVLabs released SoL-Pi, an MIT-licensed extension for the Pi coding agent that packages four efficiency mechanisms discovered by automated research loops rather than by engineers. It installs on an unmodified Pi release, and every mechanism is opt-in and off by default. The four: Action Fusio

The Stack

China's new five-year plan puts the AI model on the device

Beijing wrote its AI hardware ambitions into the 2026–2030 industrial blueprint on Tuesday, a former FTC chair told the labs they are already breakable under old law, and open-weight models went head-to-head with the frontier on real coding work. China's industry ministry and the National Development and Reform Commission issued the 15th Five-Year Plan for electronic information manufacturing on September 15, and its AI chapter is about silicon under the model, not the model itself. The plan na

The Stack

Same Qwen3 voice model, 14x faster — hosting is the product now

A YC-backed inference shop just posted numbers showing the model on the label tells you almost nothing about how it performs. Meanwhile, a music startup raised money on the opposite bet: that abundance of AI-generated audio makes human taste worth more. Nari Labs reported the top scores on Coval's voice AI leaderboards, and the sharpest number in the release isn't its own — it's what happens to the model it didn't tune. The YC-backed host serves a 1.7B Qwen3-TTS endpoint at 8.8% word error rate

The Stack

d-Matrix stacks DRAM on its chip and claims 100 TB/s

Hot Chips 2026 wrapped with a memory bet, and the funding calendar is unusually busy for a Monday. d-Matrix says stacking memory directly on the compute die beats HBM by 20x on bandwidth density. The startup's Raptor accelerator puts a TSMC N4 logic die face-to-face on top of custom 3D DRAM and claims roughly 100 TB/s per card at 32 GB — against a practical ceiling of about 20 TB/s for today's HBM4 packages, which d-Matrix attributes to limited package "beachfront" and raw physics: driving 100

See all