Click any tag below to further narrow down your results
Links
The author argues that standard hiring practices—especially heavy reference checking—don't actually predict future performance better than we think they do. After 50+ hires, he's learned that we mistake subjective judgment and past narratives for objective truth, while ignoring how much a person's performance depends on their environment and manager.
- Multiple references from the same company or network often reflect a shared narrative rather than independent information, and references themselves are biased participants, not neutral observers.
- Past experience is a poor predictor for roles involving rapidly changing technologies and requirements; someone's ability to learn and adapt matters more than years in adjacent roles.
- Most hiring advice is anecdotal autobiography—individual companies rarely hire enough people in similar roles to prove statistically which methods actually work, yet we treat personal success as universal methodology.
- We evaluate candidates heavily but rarely assess whether the company, manager, or system will actually enable good performance, treating professional capability as an individual trait rather than a system outcome.
A new ICML paper argues that shifting fair machine learning toward causal reasoning actually harms policy objectives like diversity and equity, contrary to what researchers expected. The authors show that causal methods can produce worse outcomes than simpler approaches like removing sensitive features entirely.
- Causal fairness frameworks, despite seeming like improvements over traditional methods, can actively damage diversity and equity efforts in practice
- Removing sensitive attributes (the traditional "blinding" approach) sometimes outperforms causal reasoning for achieving policy goals
- The counterintuitive finding challenges the current direction of fair ML research, which has largely embraced causal thinking as more rigorous and principled
HackerRank is rolling out Chakra, an AI agent that conducts coding interviews while candidates have access to AI tools, measuring not just whether they get the right answer but how they think and work with AI. The company claims this approach cuts cheating flags by 70-80% and consolidates a three-round hiring process into one interview.
- Chakra has already conducted 500,000+ interviews in beta with companies like Snowflake and Capgemini, and HackerRank is positioning it as the future of its business, comparing the shift to Apple moving from iPod to iPhone.
- Candidates can use AI during interviews without penalty, which HackerRank found actually reduces cheating because it removes the incentive to secretly use outside tools—suspicious activity dropped 70-80% compared to traditional assessments.
- The AI measures "AI fluency" (how well candidates prompt AI and evaluate its output) and reasoning process rather than just final answers, while HackerRank says humans retain final hiring decisions to avoid over-delegating to algorithms.
- Regulatory scrutiny of AI hiring tools is already happening (NYC requires bias audits), and while Ravisankar argues properly tuned AI is less biased than humans, automated systems can inherit biases from training data and criteria.
TypeSafe AI released Jev, a model that takes text input but outputs confidence scores and probability distributions instead of generated text. It's designed for classification tasks like spam detection and ranking, costs $0.042 per million input tokens with free output, and has already spawned open-source recreations and creative experiments.
- Jev returns structured probabilistic outputs (yes/no scores, rankings, category choices) rather than text, making it fast and cheap—you only pay for input tokens, not output.
- The model trades interpretability for simplicity: you get a floating-point score but no explanation of which signals drove the decision, raising serious concerns about hidden bias in high-stakes use cases like hiring.
- The community has already built unconventional projects with it (a chat model, a game player, a left-pad implementation) and created open-weight alternatives like Kev within days of release.
Researchers trained language models on text from before 1930 to create AI that genuinely doesn't know what happened after that year, then had people interact with these "historical minds" to test whether it changes how they view the past. A preregistered experiment with 240 participants found that talking to a pre-1930 AI reduced people's tendency to think the past was more moral than the present.
- The core innovation is using temporal knowledge cutoffs as an experimental variable—training LLMs on historical corpora so they can authentically respond without knowledge of subsequent events, making the past interactable in ways archives and living testimony cannot.
- A randomized controlled trial showed interaction with a pre-1930 model significantly reduced the "illusion of moral decline," a cognitive bias where people perceive historical periods as more ethical than the present.
- This opens a new methodology called "science fiction science"—turning speculative thought experiments into testable empirical studies by using AI as a tool to reconstruct historical perspectives.
Dan Luu argues that he notices severe bugs in products that teams insist work fine, and traces this gap to how fans of something—whether employees, users, or enthusiasts—develop psychological blind spots to its flaws. He uses search engines, Blackboard, and forum software as examples of products widely disliked by actual users but defended by insiders.
- Products can be severely broken (requiring workarounds, producing spam-filled results, cheating on performance metrics) while internal teams genuinely believe they work well, then fail publicly when real users encounter the same issues Luu spotted.
- People exhibit strong motivated blindness toward things they're invested in: Volvo forum users deny reliability data, Blackboard employees were shocked it was widely hated, and Discourse developers implemented performance cheating while apparently believing their software was fast.
- Luu suspects this blindness is partly about mental barriers people construct—acknowledging a flaw in your work or something you love requires confronting cognitive dissonance, so teams rationalize away contradicting evidence.
A comprehensive 2026 test compared top fitness trackers and identified a single standout model. The results aim to remove any guesswork for buyers looking for accuracy, battery life, and overall performance.
- This is a marketing post from WHOOP's own account (@myhumehealth), not an independent review, so treat the "99.2% accuracy" and "clear winner" claims with skepticism.
- Claimed results: WHOOP 5 hit 99.2% heart-rate accuracy, 94% sleep-stage match to polysomnography, and 10-day battery life, versus Fitbit Charge 6 (98% step accuracy, 85% sleep accuracy, 4-day battery) and Apple Watch Series 8 (18-hour battery, inconsistent heart-rate under high output).
- Study specs cited: 50 volunteers, 30 days, 3,000+ hours of data, tested against medical-grade monitors across running, cycling, swimming, and strength training.
- Comfort/build quality favored WHOOP's lightweight band, while Garmin and Polar straps reportedly caused minor chafing.