doorss

Founders are using AI-generated code to overcome longstanding bottlenecks in product development, accelerating how quickly start-ups can build and ship products.

A personal reflection on how the shift to LLM-assisted programming has disrupted the author's decades-long practice of coding to music, fundamentally changing the nature of the work.

Margaret Storey revisits the concept of 'cognitive debt' — the accumulated cost when developers don't understand code generated by AI agents — and reports on early findings about its effects on software teams.

Citadel Securities reports that job postings for software engineers are rapidly rising, countering narratives of AI-driven job displacement in the field.

A research project examining how large language models are systematically distorting written language patterns, documenting specific linguistic shifts attributable to LLM-generated text.

Andrew Kelley of the Zig project explains that the mistakes LLMs make are fundamentally different from human mistakes, making LLM-assisted contributions detectable and problematic for code quality.

An investigation into the advertising and data collection infrastructure built into modern connected cars, detailing how vehicles track and monetize driver behavior.

A security researcher discovered that Microsoft Edge keeps all saved passwords in clear text in memory, even when the user hasn't accessed them, creating a significant attack surface.

A detailed breakdown of how ChatGPT's advertising system works, including the full attribution loop from ad placement through prompt relevance matching to conversion tracking.

Amsterdam kielsi liha- ja fossiilisten polttoaineiden mainonnan. Hollannissa on tehty vastaavia päätöksiä aiemminkin.

Anthropic's discovery team shares results from benchmarking Claude's ability to perform bioinformatics research, testing whether LLMs can meaningfully contribute to biological data analysis.

Using privacy-preserving analysis, Anthropic studied how people ask Claude for personal guidance on topics like career decisions and relationships, and evaluated Claude's sycophancy in responses.

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

Bluesky published technical details about how their "For You" feed works, highlighting the platform's unique feature that allows anyone to run their own custom feed algorithm.

Bluesky experienced a prolonged DDoS attack lasting nearly a full day, testing the decentralized social platform's infrastructure resilience.

Nokian Renkaat esitteli uudenlaisen nastarenkaan, joka kuluttaa teitä vähemmän. Yle vieraili rengaslaboratoriossa.

Suomi putosi kuudenneksi lehdistönvapausindeksissä. Ennen vuotta 2016 Suomi oli ykkönen yli kymmenen kertaa.

Pieni suomalainen kunta kokeili tekoälyä ja sai kuukausien työn tehtyä päivissä. Sitran mukaan Suomi uhkaa jäädä jälkeen, koska onnistuneita kokeiluja ei osata laajentaa.