A look at "humanizer" and "autotyper" apps that help students evade AI-detection software by slowly auto-typing essays and making AI text sound less robotic (Dana Goldstein/New York Times)
THE SO WHAT
The existence of "humanizer" and "autotyper" tools is a reminder that AI-detection is already an adversarial space, not a compliance checkbox. If your policy relies on detectors rather than workflow design and incentives, assume motivated users will route around it.
READ THE SOURCE
MORE FROM THE WIRE
Applied AI‘Use GenAI.mil, do the best you can': Pentagon officials boast of using AI to generate Congress reports
When 1.5 million Pentagon workers are told to use GenAI.mil for Congressional reporting, AI-generated text becomes part of the national security record, not just internal drafts. Any org doing the same needs hard policies on provenance, retention, and human sign-off before synthetic content escapes into formal oversight channels.
Applied AISignal’s Meredith Whittaker wants you to remember that AI chatbots ‘are not your friends’
Whittaker is reframing chatbots as data-hungry services, not companions — that’s a cue for enterprises to harden policies around what employees disclose to third-party assistants. Treat every assistant interaction as if it could be logged, learned from, and subpoenaed, and adjust your governance and UX copy accordingly.
Applied AIChinese scientists ran an AI program on a virtual light-based computer system inside its real 'digital twin' PC — you can't get more meta than that (thanks Inception)
Digital twins for optical compute mean you can design, train, and debug against a simulated photonic stack before you ever touch hardware — that compresses iteration cycles and de-risks exotic accelerators. If you're betting on non-CMOS compute, start asking vendors for their simulation and tooling story, not just TOPS and watts.
Applied AIIn the Weights is your new AI-centric vanity search
AI-native reputation scores are coming for every professional graph — once there's a number, people will optimize for it. Early signal for operators: assume your team, candidates, and brand will be scored by opaque AI systems and start deciding what you actually want to be legible for.