Intel Punches Back

It’s Thursday, October ninth, and Intel is swinging for a comeback with Panther Lake, Cursor’s wild valuation chatter is heating up, an open-source challenger to DeepSeek just landed a massive check, we may be edging past attention-only models, and Microsoft is making a big healthcare push. Here’s what you missed today in the world of tech.

Intel unveiled Panther Lake, its first chip on the 18A process, promising stronger performance and better battery life than Lunar Lake while scaling from thin-and-light laptops to more powerful rigs and even handhelds. Three variants blend new P-cores and E-cores with beefier Xe3 graphics, plus smarter scheduling that shifts game workloads to efficiency cores to free power for the GPU.

Intel hammered the built-in-the-United-States message at its Arizona fab, the first slated for mass 18A production, and admitted credibility now hinges on shipping products that consumers can feel. The next node, 14A, will need outside customers at real volume, and leadership said plainly not to trust them until the results arrive.

Discord said a breach at its customer support vendor exposed approximately seventy thousand users’ government ID photos tied to age checks. The company rejected inflated claims from extortionists, contacted affected users, and is working with authorities.

Cursor’s parent, AnySphere, is weighing new funding that could mark it near a thirty billion dollar valuation on roughly five hundred million dollars in annual recurring revenue as of June, with a path to one billion by year end. It’s pushing deeper into enterprise, building its own models to cut inference costs, exploring licensing of aggregated interaction data, and even got a public nod from Nvidia’s CEO.

Reflection AI raised two billion dollars led by Nvidia at an eight billion dollar valuation to build an open-source model that can stand toe to toe with DeepSeek. The bet frames open source as a strategic need for the United States and underscores how compute and talent costs are driving mega-rounds.

Samsung introduced the Tiny Recursion Model, a seven million parameter network that uses iterative reasoning to punch above its weight on structured tasks. The code is MIT-licensed, and on problems like small-grid Sudoku, a simple recursive MLP can beat far larger attention-based models.

Microsoft is leaning into healthcare for Copilot with Harvard Medical School while it builds more independence from OpenAI, testing homegrown models and even using Anthropic in parts of Microsoft 365. It still trails ChatGPT in consumer reach, but the company is retooling fast to close the gap.