doorss

Anthropic shares results from benchmarking Claude's ability to conduct bioinformatics research, testing whether the model can solve biological data mysteries that require multi-step reasoning.

Using privacy-preserving analysis, Anthropic examined how people use Claude for personal life decisions—from career moves to relationships—revealing patterns in how users seek AI guidance on deeply human questions.

Anthropic's survey of 81,000 Claude users finds that people in roles more exposed to AI have greater concerns about AI-driven job displacement, with concerns also higher among lower-income workers.

Simon Willison documents the Zig programming language's strict ban on LLM-assisted contributions, including a quote from the project explaining that the kinds of mistakes humans make are fundamentally different from LLM errors.

Citadel Securities reports that software engineering job postings are rapidly increasing, countering narratives about AI-driven job losses in the field.

The FT argues that whether AI can perform a task is only a small part of the employment picture, and the broader 'jobpocalypse' narrative overlooks important structural and economic factors.

A post arguing against the common framing of LLMs as 'junior engineers,' examining the ways this analogy breaks down and misleads expectations about AI coding capabilities.

Surveillance company Flock accessed cameras in a children's gymnastics room to use as a sales demonstration; the city renewed the contract anyway despite learning about the incident.

A researcher demonstrates that Android devices can leak real IP addresses even when the always-on VPN setting is enabled, undermining a key privacy protection.

Researchers discovered a stable identifier in Firefox's IndexedDB that could link separate Tor browsing sessions, compromising user anonymity across supposedly isolated private identities.

A breakdown of the full attribution loop behind ChatGPT's ad serving mechanism, detailing how OpenAI tracks ad placements and measures their effectiveness within conversational AI.

TV advertising is gaining renewed value as streaming platforms introduce ad-supported tiers, offering marketers new channels as digital platforms seek additional revenue sources.

Bluesky's AT Protocol team explains how they serve the "For You" algorithmic feed, highlighting the platform's unique design where anyone can run their own custom feed implementation.

Bluesky experienced a sustained DDoS attack lasting nearly a full day, disrupting service on the growing decentralized social platform.

Perussuomalaiset käytti tekoälyllä luotua kuvaa, jossa näkyy hunnutettu lapsi koululuokassa. Tutkijan mukaan tekoälykuvat voivat hämärtää todellisuuden ja kuvitelman rajaa.

Pieni suomalaiskunta kokeili tekoälyä hallinnossaan ja sai kuukausien työn valmiiksi päivissä. Sitran mukaan Suomi uhkaa jäädä jälkeen, koska onnistuneita kokeiluja ei osata laajentaa.

Suomi sijoittui lehdistönvapausindeksissä kuudenneksi. Ennen vuotta 2016 Suomi oli ykkösmaa yli kymmenen kertaa.