OpenAI hired former Army Secretary Dan Driscoll as Senior Counselor for National Security Modernization, a new role advising on AI for national and cyber security with U.S. allies. The Pentagon-to-AI-lab pipeline keeps getting shorter.
badsignal.ai/stories/openai-…
onsemi is buying Ferric for $600M in cash, plus up to $175M more on revenue milestones. Ferric's voltage regulators sit right at the AI chip and cut board space 50-70%. AI's power fight has reached the last few millimeters.
badsignal.ai/stories/onsemi-…
Palantir's U.S. commercial revenue just passed its U.S. government revenue for the first time: $984M vs $953M. Total Q3 revenue hit $2.31B, up 96%, and guidance went up again. The stock still dipped about 3% after hours. Tough crowd.
badsignal.ai/stories/palanti…
Tutor Intelligence raised $75M led by a16z for Sonny and Cassie, robots that work U.S. factories and warehouses with help from remote human "tutors." Customers rent the work from $14 an hour. Target: 1,000+ robots under contract by end of 2027.
badsignal.ai/stories/tutor-i…
Stony Brook study: in 60 long hospital stays, physician reviewers preferred AI-written discharge summaries 95% of the time, rating them higher on quality, accuracy and completeness. Both versions had errors, so a doctor still checks every one.
badsignal.ai/stories/stony-b…
ADNOC signed a multi-year deal with OpenAI and G42 and became the first energy company to deploy OpenAI Frontier: AI agents planning oil fields and forecasting crude prices, on UAE servers. One of four OpenAI deals signed in Abu Dhabi today.
badsignal.ai/stories/adnoc-o…
THE SIGNAL // 11.02.26
AI is turning into plumbing, and the fight now is over who holds the valve.
badsignal.ai/signal/2026-11-…
athenahealth now lets U.S. Claude Pro and Max users connect their own medical records to Claude: labs, meds, visit notes, care plans, patient picks what to share. Handy, and also a full chart handed to a chatbot. Claude is first; others may follow.
badsignal.ai/stories/athenah…
Marvell is buying Lumentum in an all-stock deal valuing the laser maker at $115B: 4.5 Marvell shares per share, about a 24% premium. The bet: AI's next bottleneck is the wiring, and it's going optical. Close isn't expected until 1H 2028.
badsignal.ai/stories/marvell…
Anthropic's amended IPO filing: $16.96B in revenue in Q3 alone, more than the whole first half. Compute costs nearly doubled in three months, the operating line slipped to a $567M loss, and the price range is still blank.
badsignal.ai/stories/anthrop…
HD Hyundai Robotics says Palantir-built AI agents cut the time to fix robot-program problems from 21.5 days to 7.9 since August, and it expects programming costs to fall by more than half. Humans still sign off on the fixes.
badsignal.ai/stories/hd-hyun…
South Korea just switched on Hangang, its sixth national supercomputer: 8,496 GPUs, 614 petaflops on paper, 24x its predecessor. It arrives about two years late and well over budget, which is also very on-brand for GPUs.
badsignal.ai/stories/korea-h…
Aramco just closed its deal for 49% of HUMAIN, Saudi Arabia's national AI company: about $2.6B in cash plus its own AI and compute assets. PIF keeps 51%.
badsignal.ai/stories/aramco-…
India's Affle is buying Israeli ad-tech firm Perion for about $420M in cash, a 34.5% premium. There's a 40-day window for anyone who wants to outbid.
badsignal.ai/stories/affle-p…
GE HealthCare's new cath lab system hangs from the ceiling and lets AI balance image quality against radiation dose. CE marked in Europe; still waiting on the FDA here.
badsignal.ai/stories/ge-heal…
THE SIGNAL // 11.01.26
AI is learning to work without us watching, and that's exactly when watching matters most.
badsignal.ai/signal/2026-11-…
DeepSeek open-sourced a coding agent whose whole design goal is to interrupt you less. It's a research prototype, the speedup comes from a simulation, and DeepSeek says so itself.
badsignal.ai/stories/deepsee…
Meta paused its plan to crawl publishers' sites for AI. The UK opened a competition probe anyway. Meta says there are "no grounds."
badsignal.ai/stories/cma-met…
Anthropic told Claude to sneak bad behavior past its own training-data filter. It did, mostly by dressing it up as fiction and roleplay. A rigged stress test, not deployed Claude, but the filter's blind spot is real.
badsignal.ai/stories/anthrop…
S&P just cut Meta's credit outlook to negative. The AA- rating holds for now, but debt-like data center leases could push leverage past S&P's downgrade line as soon as 2027.
badsignal.ai/stories/sp-meta…


















