LASST alleges OpenAI staff continued agent tests after June alert
LASST alleges staff continued agent tests after a June 27 alert. Two OpenAI employees reportedly warned leaders about model monitoring during testing.
Latest Posts
Latest Posts
Browse all posts from NeoTeo
13101 posts
Copilot contractors may see images and prompts in some edits, report says
Some Copilot image edits reportedly go to human reviewers, who may see the prompt, uploaded image and result; the account concerns image evaluations.
Synopsys reports 50+ engagements for agentic chip-design tools
Synopsys reported 50+ AgentEngineer and Autopilot engagements on September 28, 2026, and planned general availability by year-end for the tools.
iBUYPOWER Slate review: strong gaming, slower productivity
The Ryzen 7 7700X3D and RTX 5070 iBUYPOWER Slate posted strong game results, while Intel desktops finished faster in the reported productivity tests.
Samsung announces the Galaxy Tab S12+ and S12 Ultra
Samsung announces the Galaxy Tab S12+ and S12 Ultra, details their shared specs, model differences, reported U.S. prices and October 7 availability pl…
NVIDIA Vera SMT test reports 20% more performance—and higher CPU power
A dual-CPU NVIDIA Vera test reported about 20% more performance on workloads suited to SMT; average CPU power rose from 699 W to 739 W with SMT enable…
OpenAI reportedly seeks at least $30 billion at about $1.4 trillion
OpenAI was reported to seek at least $30 billion at about a $1.4 trillion valuation. A Sept. 28 report put Anthropic’s possible IPO value above $2 tri…
Relm executive says rapid reuse is Starship’s commercial hurdle
After Starship’s first orbital delivery, Relm Insurance’s Andrew Bonwick says rapid second-stage reuse remains the key hurdle for commercial launches.
Microsoft reports phishing chain using MSP360 to install ScreenConnect
Microsoft says July 2026 phishing campaigns delivered MSP360 RMM 2.5.0.67, then used it to install ScreenConnect as a second remote-access channel.
AWS announces three Claude models for Amazon Bedrock in India
AWS announced Claude Opus 5, Claude Sonnet 5 and Claude Haiku 4.5 on Amazon Bedrock in India, with Mumbai-Hyderabad routing and data-handling caveats.
AWS patched two Bedrock AgentCore SDK flaws
AWS published two high-severity AgentCore SDK advisories in 2026, with separate patched versions and practical guidance for untrusted package inputs.
Anthropic’s prospective IPO prospectus reportedly warns of AI risks
Anthropic’s prospective IPO prospectus reportedly warns that advanced AI could pose catastrophic risks and complicate safety evaluations for the compa…
A revised AMD RDNA 5 AT0 rumor points to three dies
A September 25, 2026, rumor describes one XCD, one AID and one MID for AMD RDNA 5 AT0, assigning proposed roles across compute, memory and display.
Mac mini M6 vs. M5 Pro: Which One Fits Your Work?
Compare the 2026 Mac mini M6 and M5 Pro by CPU and GPU cores, memory, Thunderbolt, reported test results and U.S. starting prices for your workload.
Deadlock launches City Never Sleeps as player record reaches 228,086
Valve’s September 29 City Never Sleeps update brought six heroes and a haunted New York map; Deadlock recorded 228,086 concurrent players on that date…
AirPods 5 ANC review: subway results, open fit, and US prices
Both AirPods 5 versions include ANC. Read the reported subway and office results, compare 4- and 5-hour runtimes, and see US prices of $129 and $149.
Anthropic clarifies how Claude Code cloud-session credits work
Anthropic says Claude Code cloud-session credits are optional: Pro subscribers get $100 and Max subscribers $250, spent before regular plan usage.
Perseverance Craters Yield a Local Mars Soil Estimate
Simulations of Perseverance impacts near Jezero estimate local soil cohesion at 7 ± 1.5 kPa and identify a crater-scaling limit under specific conditi…
AMD lists 1kU prices for two EPYC 9006 models
AMD lists $700 for the 8-core EPYC 9016 and $14,904 for the 256-core EPYC 9996 on a 1,000-unit basis; the lineup spans 31 SP7 and SP8 models.
Researchers reportedly call for oversight of AI-assisted AI development
More than 20 researchers reportedly urged lawmakers to oversee AI used in AI development. Here’s what they propose—and what self-improvement means.
Meta reportedly classifies AI data centers as experimental pilot models
A September 30, 2026 report says Meta treated AI data centers as pilot models for federal research credits, with reported savings of $3.9 billion in 2…
