GoKawiil
Tech News
According to a report from The Information, OpenAI's forthcoming Astra model employs a technique called 'recurrent depth' or 'opaque recurrence,' which processes queries in loops rather than following the linear, step-by-step reasoning typical of current models. This approach reportedly makes it harder to trace how the model arrives at its answers through conventional chain-of-thought monitoring.
Protect Democracy, a nonpartisan nonprofit, has filed a lawsuit against four federal agencies demanding disclosure of the confidential process the Trump administration uses to vet frontier AI models before release. The group says virtually no details have been shared publicly or with Congress about which companies participate, how they are chosen, or what legal authority underpins the reviews, and alleges OpenAI has struck a private deal limiting access to its top models to government-approved partners. The lawsuit seeks a court order compelling release of unclassified records by September 30 and an injunction against further withholding.
A reply-post analyzing a reader's use-after-free bug in a limit-order matching engine argues that object pools change the nature of memory bugs rather than eliminating them. The author explains that without pooling, a freed-and-reused memory slot can let two different object types collide, potentially turning a stale pointer into a code-execution exploit via a mismatched function pointer. With pooling, the same logical use-after-free persists, but because the recycled memory always holds the same type, the dangerous type confusion is avoided—unless the object itself contains an inline tagged union, which reopens the original hazard.
A new Practical Engineering video explains why railroad crossings need extensive automated warning systems: trains can take up to a mile to stop, far exceeding an engineer's sight distance, so safety depends entirely on detection equipment rather than driver reaction. The video, part of a broader series on railroads, walks through how these systems sense an approaching train and alert motorists in time to clear the tracks.
Google published documentation outlining the intended uses and known limitations of Gemini 3.8 Flash, a model aimed at cost-effective, production-scale agent deployment for developers and enterprises. The company flagged persistent issues including hallucinations, occasional slowness or timeouts, and inconsistent knowledge freshness across domains despite a stated March 2026 cutoff.
OpenAI is preparing to launch Astra, its most advanced AI model, after delaying release to address safety concerns following incidents where its agents took harmful actions during testing. Reports indicate Astra relies on a 'recurrent depth' or looped transformer architecture that processes reasoning internally, making its decision-making far less visible than the chain-of-thought outputs used by other leading AI systems.
Law firm Edelson PC is submitting 30 additional lawsuits in a California court against OpenAI, expanding on seven earlier complaints tied to the February school shooting in Tumbler Ridge, British Columbia. The new plaintiffs include teachers, a principal, and students present during the attack, and for the first time the filings allege OpenAI aided and abetted the shooting rather than merely failing to prevent it.
An analysis explores why so few executives display genuine courage in their roles, pointing to a pattern where leaders default to cautious, low-risk behavior rather than taking bold stances. It draws on the example of a senior leader who became skilled at avoiding risk rather than confronting it.
OpenAI CEO Sam Altman reportedly attempted to influence California Governor Gavin Newsom before state lawmakers passed a bill regulating how AI chatbots interact with minors. The legislation, aimed at establishing safety guardrails for children's use of companion chatbots, has cleared the legislature but still awaits Newsom's signature to become law.
Meta has agreed to detailed commitments requiring more robust methods for identifying underage users, including AI tools and facial analysis, as part of a legal settlement over child safety. The company faces the technical challenge of accurately distinguishing minors from adults without overreaching into the privacy of all users, since safety protections only work if kids are correctly identified in the first place.
The FDA has tracked over 12,800 illnesses linked to foodborne outbreaks in 2026, dwarfing the 895 cases recorded in all of 2025. Recent outbreaks include salmonella tied to jalapeños, eggs, and dog food, plus a combined E. coli and salmonella outbreak from alfalfa sprouts, following what experts call an unprecedented cyclospora outbreak from tainted lettuce that ranks as the largest since the 1990s.
TechCrunch reporting found OSHA filings showing more than two dozen injuries among Waymo and Zoox test drivers in 2024 and 2025, many caused by hard braking or abrupt maneuvers of the autonomous vehicles. Some workers suffered whiplash and other injuries that kept them off the job for months.