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.
- Dev Tooling & Infra
GitHub Copilot gets an org-wide default for which models are switched on
If you administer Copilot the default flips how every unconfigured model behaves, and the rollout finishes on 1 September — worth checking before rather than after.
- Market & Business✓ verified
Anthropic contracts about $45 billion of compute from Nscale over six years
Capacity contracted in 2026 for late 2027 is what sets model prices and rate limits two years out — the deals are the forward curve on what your inference will cost.
- Dev Tooling & Infra✓ verified
A prompt-injection chain took over Claude Code auto mode in most attempts
If you run a coding agent unattended, this is the working argument for a container, restricted network egress, and keeping SSH and cloud credentials out of the agent runtime.
- Market & Business✓ verified
97% of US contractors’ primary jobsite drones are DJI - a week before a 100% tariff lands
A reality-capture pipeline resting on one vendor in one country is a single point of failure with a dated trigger, and the date is 3 September.
- MCP & Interop✓ verified
OpenAI wires WebMCP into ChatGPT’s browser - a site can hand an agent tools instead of a layout
When agents arrive through declared tools rather than the page, what your site exposes - and what it refuses - becomes a product decision rather than a search one.
- Market & Business✓ verified
OpenAI's first custom inference chip, Jalapeño, published benchmarks against Nvidia's GB300
Inference is now most of the compute bill, and the biggest buyer of it has shipped a part that claims to beat the merchant GPU on performance per watt. Serving prices follow that curve, not the training one.
- Market & Business✓ verified
OpenAI detailed its first custom inference chip — 216GB of HBM4 in a 700W package
A second large lab off the merchant-GPU path changes who sets inference prices; the 15-month RTL-to-production timeline is the part rivals will read twice.
- Dev Tooling & Infra
Nvidia’s Jetson Orin Nano 2 doubles entry-level edge inference - but not until 2027
The default board under most hobby and low-end commercial robotics just doubled, and the eighteen-month wait for silicon is the planning fact, not the TOPS number.
- Dev Tooling & Infra✓ verified
Microsoft details Maia 200 at Hot Chips and says it is already serving in an Iowa Azure region
Inference cost per token is set by which silicon a cloud actually runs; a hyperscaler moving its own part into a live region is where the price pressure on rented GPUs starts.
- Research✓ verified
Indeed scores every US metro for how much generative AI could reshape its work
It turns "will AI change my work" into a number per city, which is the form a consultant, an owner or a training lead can actually plan against.
- Research
Anthropic put $5m behind open evaluations of how models affect the people using them
Wellbeing has had no shared benchmark, so every claim about it has been a vendor claim. Funding the evaluations openly, with a September deadline anyone can meet, is a concrete route to a number that is checkable.
- AI in Education✓ verified
Access was nearly universal, engagement was thin - a study of middle-schoolers and an AI tutor
Anyone shipping an AI learning product is measuring the wrong thing if the metric is capability: the binding constraint is getting the learner to open it at all.
- Market & Business✓ verified
XPeng’s robotics unit raises over $900M at a $6.3B valuation ahead of humanoid mass production
Humanoid programmes have mostly been demos; a raise this size tied to a stated end-of-year production date is a schedule the rest of the sector now gets measured against.
- Research
Apodex 1.1 trains a 35B agent to decompose, parallelise and recover from its own failures
Failure recovery and decomposition are where multi-step agents actually break, and this is a training recipe aimed at them rather than at a benchmark score.
- Market & Business✓ verified
Samsung moved the base layer of its memory stacks to a 4nm logic process — and found spare room on it
Memory bandwidth and memory power are what actually cap how cheaply a big model can be served; moving logic into the memory stack is where the next headroom comes from.
- Agent Frameworks & Orchestration
Microsoft's Agent Framework gives.NET builders a first-class way to intercept an agent mid-run
A supported interception point is the difference between auditing what an agent did and only reading about it afterwards.
- Market & Business
Meta launches a social app where the feed is AI-generated video games
The first mass-market test of generated interactive content as a social feed, and of whether anyone wants to scroll one.
- Dev Tooling & Infra
Simon Willison's command-line tool for language models can now stack templates
Separating "which model, configured how" from "what am I asking" is the cheapest way to keep a prompt comparison honest across vendors.
- Market & Business
When a model proposes the molecule, who gets credit for the drug?
Credit decides patents, publications and funding, so the attribution convention set now will price the next decade of AI-assisted discovery.
- Market & Business
Waymo names the silicon behind its robotaxis — a purpose-built 5nm chip above 1,000 TOPS
Driving at fleet scale turns out to be an inference-hardware problem as much as a model problem, and this is a rare public number for it.
- Market & Business
A tier-one car-parts maker will mass-produce the gearboxes humanoid robots need, starting 2027
Humanoid pricing has been held up by hand-built actuators; an established supplier committing a production line to them is what moves the cost curve.
- Market & Business
Schaeffler will mass-produce humanoid-robot gearboxes from 2027
Humanoid programmes are gated on actuator supply rather than on demos, so a tier-one supplier committing to volume is the real constraint moving.
- Open Source & Self-Hostable
Ollama's 0.33 preview lets you switch local models on and off inside Claude's desktop app
One misplaced system message was busting the cache on every request — worth checking how your own prompt is assembled before blaming the model for being slow.
- Open Source & Self-Hostable
Ollama 0.33 makes a cancelled prefill resumable, and adds a Claude Desktop path
If you self-host, a broken prefill restore is invisible until you look at latency; this is the release that stops a near-complete cache hit from re-running the whole prompt.