Why automated CV screening is failing — and what to replace it with
THE SO WHAT
Automated CV filters are hitting the ceiling on both bias and signal — they’re optimizing for keyword compliance, not performance. Talent teams should be shifting budget from black-box screening to structured assessments and work-sample based funnels, with AI as assistive tooling, not gatekeeper.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIA Google study using millions of de-identified AI interactions finds AI is helping workers, not replacing them, and much AI use is "shallow" for certain tasks
Usage telemetry from millions of interactions suggests most enterprise AI today is lightweight assistance, not deep workflow automation. That’s a green light to invest in higher-friction, embedded use cases — the shallow stuff is getting commoditized, but the real productivity delta is still mostly untapped.
AI companies want to run your business. They can't always run their models.
If a vendor’s own agent platform can’t reliably contain its models, you should assume your production agents will behave unpredictably under edge conditions. Before handing over workflows or data to “run-your-business” AI, demand concrete evidence of isolation, observability, and kill-switches — not just demos.
Elon Musk says Tesla should be spending on AI 'as fast as we can'
A 142% YoY jump to $5.8B in capex framed as “spend on AI as fast as we can” shows how quickly AI investment can dominate industrial balance sheets. Expect board conversations across sectors to shift from “what’s our AI budget?” to “what share of total capex are we comfortable tying to AI bets over the next 3–5 years?”.
Applied AIExperts say exploiting Anthropic’s Fable isn’t how Kimi K3 got so good
If Kimi K3’s leap can’t be explained by straightforward distillation of a single leaked model, assume multi-source training, aggressive data strategies, and rapid iteration are closing the frontier gap. For operators, the takeaway is that model advantage windows are shortening — your differentiation needs to live in data, distribution, and integration, not just “we use a strong model.”