Research
Everything Intel keeps, in one place — every brief we run, every department, readers' own, and every finding behind them. Browse below, or see the same body of work as a connected graph on the live map.
All coverage
Everything Intel has read, newest first. Each title opens at its original publisher.
- Open Source & Self-Hostable
Ollama 0.33.3 lands, adding image and audio input for gemma4
Multimodal input on a local runner means image and audio prototypes no longer have to leave your machine.
- Market & Business
NVIDIA agrees to acquire Hugging Face
The neutral hub most teams pull open models from is being bought by the company that sells the hardware to run them — worth watching if your stack assumes it stays vendor-neutral.
- Dev Tooling & Infra
GitHub CLI's Linux signing key expires today — APT and RPM installs break until you rotate it
Any Linux box or CI image that installs gh from the official repo stops updating this weekend unless the new key is added — a silent breakage in build pipelines.
- Agent Frameworks & Orchestration
A practical pattern for giving coding agents a memory you control
Agent memory is the difference between re-explaining your codebase every session and not — and owning it keeps it portable across tools.
- Research
A 350M model taught to emit reliable structured output in 100 GRPO steps
Reliable JSON out of a tiny local model is the cheapest fix for the parsing failures that break agent tool calls.
- Dev Tooling & Infra
VS Code 1.136 adds an agent that works a pull request until it is ready to merge
The last mile of a pull request — rerunning checks, clearing conflicts, answering review notes — is the part that actually eats an afternoon.
- Market & Business✓ verified
Take data centers out and US nonresidential construction spending sits at a three-year low
The AI build-out is now the only thing holding up an entire category of construction demand — an opening if you sell into it, a concentration risk if you depend on it.
- Open Source & Self-Hostable✓ verified
NVIDIA opened a proxy that lets one application speak both the OpenAI and Anthropic APIs
Changing model or provider is normally an application rewrite; a translating proxy turns it into a routing rule you can measure both sides of.
- AI in Education✓ verified
New York City is set to bar student-facing AI tools through eighth grade
Anyone building or selling a learning product now has to say which side of a grade-level line it sits on, and the biggest district in the country just drew one.
- Frontier Models✓ verified
Google's Gemini 3.8 Flash holds the old price and adds a cybersecurity-only sibling
Third Flash release in six weeks at the same rate card — the cheap tier is where most production traffic actually runs, and working harder per task is a cost change even when the price is not.
- Agent Frameworks & Orchestration
WeChat Pay opens its agent payment card to DeepSeek Harness and OpenClaw
Agent-initiated payment is shipping at consumer scale with a segregated balance and a human confirmation step — a concrete answer to "how do you let an agent spend money".
- Open Source & Self-Hostable
Hugging Face published 200-plus WebGPU kernels so models can run in the browser
Inference in the browser is the cheapest deployment there is — no server and no per-token bill — and fast GPU operations across mismatched devices have been the missing floor under it.
- Dev Tooling & Infra
GitHub Copilot can now approve a pull request, if an admin switches it on
If your merge rule counts approvals, this is the first setting under which a machine can satisfy it — worth deciding on purpose rather than finding out during a release.
- Research
Gemini gains agentic video understanding, with lower token cost
Video has been the expensive modality; a cheaper token path makes video-in-the-loop features worth costing out again.
- Frontier Models✓ verified
Anthropic released Claude Fable 5.1, priced about 25% below Fable 5 for typical work
The change that shows up on a bill is the cache-read price, which is where long agentic runs spend; Devin's team said it is what finally made a Fable-class model economical for their code review.
- Dev Tooling & Infra
Vercel adds per-person spending caps to its AI Gateway
Vercel names the case out loud: one person’s unsupervised coding agent could previously drain a shared team budget, and now it cannot.
- Market & Business
Companies started ranking staff by token spend, and the backlash has a name
The figures come from one vendor's reporting on other companies' internal boards, but the underlying question is now a live budgeting argument in real organisations: what number are you actually managing when you manage AI usage?
- Research
Skild AI unveils S1, a robot foundation model that picks up a task from one video
One-video, in-context task acquisition is the demo the whole robot-foundation-model field is chasing — and no benchmark score or deployment count has been published against it yet.
- Market & Business
Reframe Systems raises $40M to scale robotic microfactories that build homes
An applied case with real output numbers — ten finished homes, eight occupied, a named annual capacity — rather than a robotics demo reel.
- Agent Frameworks & Orchestration
A working checklist for agents that have to survive longer than one call
The single line worth stealing: write down the done condition before the agent starts, and check it with something deterministic rather than another model.
- Dev Tooling & Infra✓ verified
Gradium’s new default voice model reads an order number back correctly 81% of the time
Because the eval set is open, you can re-run this against your own scripts instead of taking a vendor chart on trust.
- Research✓ verified
Google’s TimesFM-3 forecasts several related series at once, with no fine-tuning
Check the licence before you plan on it: the code is Apache-2.0 but the weights are non-commercial and non-production, which rules out most business forecasting.
- Research
Anthropic moved about 150 product engineers onto security and set rules for outside cyber testers
If you get early access to a model with safeguards turned down, you are now expected to run it in a hardened, monitored sandbox — the testing-side obligations are being written down.
- Dev Tooling & Infra
XCENA and Samsung ship a CXL card that is 2TB of memory and 3,072 RISC-V cores
Memory capacity, not FLOPS, is what caps how large a model a single box can serve — and CXL cards with their own compute are the cheapest route to more of it.