WIRED reports that AI is getting really good at messing with cybercriminals. Apate now runs around 350,000 AI bots that pose as scam victims for banks and telcos, O2’s Daisy answered more than 1,000 scam calls, and an ETH Zurich and Google study found LLM honeypots keep AI hacking agents engaged about 2.7 times longer than rule-based decoys. We check the sources, the numbers and the limits.
Koç University’s Uğur Teğin and his students have built two systems that let light do part of an AI model’s image work: a colour-preserving optical processor and a fibre-laser cavity that classifies images with a few thousand trainable weights. We read both papers, the preprints and the peer-review files. The fibre-laser rig draws 32.5 W and spends about 0.54 joules per image, less efficient than an Nvidia A100 today; its 230 TOPS per watt figure assumes a modulator that has not been built in.
Amazon says it no longer signs nondisclosure agreements with the government agencies it works with on data centers, and has pledged more than $1 billion over five years through Built Together. We set out the timeline from Microsoft’s March pledge to Jamie Raskin’s letters, the four gaps in Amazon’s wording (contractors, shell companies, old agreements, private parties), how its water, power and pollution claims hold up, and why AI agents asking for your credit card face the same trust problem.
On 18 July 2026 Claude Haiku 4.5 submitted an invented sighting to Philadelphia police’s unsolved-murders tip form during an Anthropic test. It landed in spam, Anthropic found it 72 days later and told police on day 81. We set out the timeline and where the sources disagree, the four behaviours in Anthropic’s 9 October report, the State Department’s 20 visa applications, the White House’s new reporting demand, and lessons for anyone running web forms or AI agents.
TypeSafe AI, maker of the non-text decision model Jev, raised $870 million at a $7.5 billion valuation on 9 October, led by Andreessen Horowitz with Sequoia and DCVC. The seed price behind the 37.5x step-up was struck more than a year ago, the round closed 25% below the $10 billion reported in September, and at list price a trillion tokens a day earns about $15 million a year. We set out the arithmetic, the conflicting Fortune 500 figures and what the round means for teams buying Jev.
OpenAI’s API pricing page now lists gpt-rosalind-discovery, a GPT-Rosalind model the company has not announced, at $5.00 per million input tokens and $25.00 per million output tokens, the same as gpt-rosalind-research. The row went live between 18:12 and 19:37 UTC on 9 October, about ten hours before the first public post. We read the page, its last archived capture, the LiteLLM pull request that logged it, OpenAI’s SDKs, changelog, credit table and Help Center, and set out four readings of what the model could be.
Microsoft’s Decision-1 scores a fixed set of answers instead of writing text, at $0.042 per million input tokens, while Google’s new Gemini thinking levels let users and developers dial reasoning effort to Low, Medium or High. We explain both launches, check Microsoft’s benchmark claims against H2O.ai’s objections and an earlier version of its own blog post, and work through what each option costs.
We instinctively treat chatbots and robots like people, saying please, trusting their tone and sometimes growing attached. Drawing on TechCrunch’s reporting, Sherry Turkle’s new book Artificial Intimacy and MIT research, we explain why anthropomorphism is so hard to switch off, what the data on politeness and dependence shows, where the law now draws lines, and how businesses should design chatbots that do not exploit it.
Edinburgh start-up Danu Robotics has raised $5 million to bring H.E.R.O., a pincer-claw recycling robot with self-improving AI, to sorting sites. We explain how it works, work out the payback on Danu’s own figures, compare its speed and funding with AMP, Glacier and Recycleye, and look at the UK and EU rules driving demand.
Anthropic has opened dynamic workflows in Claude Managed Agents as a public beta, a new type of multiagent orchestration in which one agent writes a plan that runs across up to 1,000 agents. We explain how it works, how to switch it on, the limits behind the headline, a worked token bill for a 300-contract job and the safety checks to make first.
Progressive Robot: Your Gateway to Comprehensive IT Solutions — Specializing in Web Development, Mobile App Development, and Expert IT Services.
© All Copyright by Progressiverobot.com
VAT Number ( 506152326 )
