The Stack
AMD used its IFA 2026 keynote in Berlin to take a direct swing at Nvidia's DGX Station, and the pitch was unusually blunt about what the fight is actually over: memory. AMD unveiled the Threadripper Halo Station, a deskside workstation that pairs a 96-core Threadripper PRO 9995WX with a pair of liquid-cooled Instinct MI350P accelerators and 2TB of DDR5 system memory.
The Stack
An embedding is a list of numbers that represents the meaning of something — a word, a sentence, a photo, a song — so a computer can compare meanings using arithmetic. The rule is simple and surprisingly powerful: things that mean similar things get similar lists of numbers. That's the whole idea, and it's the foundation of semantic search, recommendation feeds, and most of what people call "AI memory." Why it matters right now Embeddings are the invisible layer under nearly every AI feature…
The Stack
Two stories that meet on the same question: what machines can now do on their own, and how badly the telling of it gets mangled on the way to your feed. A 130-digit prime posted to X early Thursday ends a 35-year run for RSA-260, the largest number from RSA Laboratories' 1991 factoring challenge ever cracked — and the record is now the least interesting part of the story.
The Guardrails
Two fresh polls, a Financial Times dispatch from the state, and a governor's course-correction all point at the same story: AI data centers have stopped being a zoning story and started being a ballot-box story in Pennsylvania, and Kimi K3 just topped a new domain-specific reasoning benchmark aimed at geoscience. Two new Pennsylvania polls put data centers at the top of voter concerns, and the issue is now reshaping both the governor's race and the down-ballot legislative map.
The Stack
If you've ever wondered why an LLM feels slow to type out long answers — or why Nvidia's engineers spent a September 2026 blog post telling model designers how many attention groups to use — the answer is a technique called speculative decoding. Speculative decoding is an inference speedup where a small, fast "draft" model guesses several upcoming tokens, and the big target model verifies them all in a single pass instead of generating one token at a time.
The Stack
OpenAI's Daybreak models now sit inside two very different products that shipped within hours of each other. Cloudflare is putting them to work patching vulnerabilities at the network edge; Proofpoint is using them to run investigations. Neither one is allowed to act alone. Cloudflare launched Vulnerability Discovery and Remediation, a service that pairs OpenAI's Daybreak cyber models — including GPT-5.6 Cyber — with traffic from its own network to find, block, and draft fixes for code flaws.
The Arena
The AI buildout's capital markets didn't slow down after hours on Thursday: an AI data center operator quietly tripled its valuation, a smart ring maker filed to go public with real profits, and an autonomous trucking company took a third swing at a listing. Crusoe has closed a funding round of more than $3 billion at a post-money valuation of roughly $30 billion.
The Arena
The trading firm's second enormous compute deal in a month, plus a forecasting model that beats everything and can't be used commercially. Jane Street signed a five-year cloud contract with Crusoe worth about $13 billion, its second huge GPU commitment in weeks. Bloomberg reported the deal, which gives the trading firm clusters of GPUs and supporting infrastructure for AI training and inference through Crusoe's cloud platform, and Reuters confirmed the report.
The Everyday
Two smart-ring stories landed within hours of each other on Thursday, and read together they describe the same pivot: the ring is done being a sensor, and everyone is racing to decide what it becomes next. Qualcomm Ventures has joined a $70 million round in Ultrahuman, valuing the Bengaluru company at about $365 million — and the real prize in the deal is silicon, not money.
The Stack
Two of today's stories point the same direction — the work is moving off someone else's server. Nvidia wants the spare machines on your network doing it, Google wants you talking to your inbox instead of typing at it, and a separate team quietly finished the largest brain map ever assembled. Nvidia released PAIR, a free tool that turns the computers you already own into one shared local AI machine.
The Stack
A token is the chunk of text an AI model actually reads and writes — not a word, not a character, but a piece of a word: "un", "believ", "able". Every language model, from the smallest phone-side model to the biggest frontier system, does its work by predicting one token after another. If you understand tokens, you understand why AI bills look the way they do, why some languages cost more than others, and why a model can count the letters in "strawberry" and still get it wrong. It is also the…
Quick Hits
An afternoon sweep: Europe's biggest AI supercomputer deal lands at €387.8M on AMD silicon, two Chinese provinces push tokens to metered-utility status, and a small devblog argues the cleanest agent memory is the one you stop scoring. Europe locks in LUMI-AI — a €387.8M, AMD-MI430X supercomputer for Finland.