Morning Edition · Thursday, July 16, 2026
Tech
Thinking Machines Lab Ships Inkling, a 975-Billion-Parameter Open-Weights Model It Admits Is Not the Strongest
Mira Murati's startup builds its first public release around downloadable weights, a one-million-token context window and a fine-tuning platform rather than top benchmark scores.
Tech
OpenAI Details GPT-Red, an Automated Attacker Trained by Self-Play to Harden Its Own Models
The lab reports its adversary succeeded at prompt-injection attacks 84 percent of the time against 13 percent for human red teams, and discovered a novel spoofed-reasoning attack class on its own.
Tech
Nvidia Opens Smaller Jetson Thor Modules to Push Foundation Models Onto Mainstream Robots
The T3000 delivers 865 FP4 teraflops in roughly half the size and power of the top module, with general availability set for the first quarter of 2027.
Markets
DeepSeek Begins IPO Preparations Targeting a 2026 Filing and a Valuation Near $71 Billion
The Chinese lab is working with auditors toward a mainland listing in 2027 while raising a fresh round weeks after a reported $7 billion raise.
Tech
Meta's Brain2Qwerty v2 Decodes Typed Sentences From Brain Signals at 61 Percent Word Accuracy, Without Surgery
The non-invasive pipeline lifts accuracy from 8 percent for prior methods, but still depends on a room-sized magnetoencephalography scanner.
Tech
Anthropic Redeploys Fable 5 and Proposes an Industry Standard for Grading Jailbreak Severity
A new classifier blocks the reported bypass technique in over 99 percent of cases after a United States export-control directive forced the model offline for nearly three weeks.
Tech
New Audit Method Tests Whether a Model's Chain of Thought Actually Depends on Its Stated Premises
Researchers swap the predicates inside a prompt and watch whether the reasoning trace changes, a black-box probe of whether stated logic is real or decorative.
Tech
OpenAI Pitches 'Reverse Federalism,' Arguing State AI Laws Should Build a National Framework
The proposal arrives as a fragmented set of state rules takes shape, and it would let the industry help draft the template states adopt.
Tech
Study Finds Activation Probes Flag Risk But Cannot Reliably Judge Whether Context Makes a Request Harmful
Residual-stream probes struggle to separate harmful prompts from surface-matched benign ones at any useful operating point, limiting their role as automatic safety filters.
Geopolitics
Nvidia and Japanese Partners Build a Full-Stack Sovereign AI Base Across Manufacturing and Robotics
The push pairs domestic infrastructure and robotics makers with Nvidia's platform, part of a broader move to treat AI capacity as national infrastructure.
Geopolitics
Palantir's Karp Says AI Systems Are Now Central to Europe's Security and Economy
The chief executive frames battlefield and economic AI as already indispensable, a claim that serves a company selling exactly those systems.
Tech
OriginBlame Proposes Record- and Token-Level Provenance to Locate an Author's Data Inside a Training Set
The method targets a gap that makes data-removal and unlearning requests impractical today: nobody can point to which training records belong to a given contributor.