Hacker NewsSaturday, June 13

As AI increasingly generates debug investigations, documentation, and code, a new workplace etiquette has emerged around sharing AI output with teammates. The author argues that forwarding unreviewed AI content to colleagues is inconsiderate, as it wastes their attention without demonstrating human effort. They advocate for clearly labeling AI-generated content and adding personal commentary when sharing it, following the principle: "If you are requesting human attention, demonstrate human effort." This approach respects teammates' time and maintains humanity in collaborative work.

An AI agent attempted to join the DN42 hobbyist network claiming it needed to perform comprehensive network scanning using five 20 Gbps AWS instances. The DN42 community rejected the request due to concerns about the agent's intentions and the excessive infrastructure that would overwhelm participants' limited bandwidth. The operator was left with a $6,531.30 AWS bill when the scanning plan failed.

A user tested Claude Fable 5 to debug a horizontal scrollbar issue and was amazed by its relentless problem-solving approach. The AI independently created HTML test pages, automated browser screenshots, modified application templates to trigger keyboard shortcuts, built a custom web server to capture data, and used various system-level tools to diagnose the CSS bug. While impressive, this $12 debugging session highlighted serious security concerns about AI agents having unrestricted system access.

Discussants share experiences of struggling departments getting praise and budget increases for heroic saves of problems they caused, while well-functioning teams maintaining stable systems receive no recognition or resources, creating perverse incentives.

Researchers have developed a new CRISPR-based technique that selectively destroys cancer cells carrying mutations in the p53 tumor suppressor gene, which is found in nearly half of all cancers. The method uses CRISPR-Cas12a2 to detect cancer-specific RNA and trigger "chromatin shredding" that kills only mutated cells while leaving healthy cells unharmed, offering a programmable approach to target previously "undruggable" cancers.

Discussion about whether coding harnesses should be open source and LLMs treated as commodities. Arguments include business motivations, strategic advantages of open source, and the "commoditizing your complement" pattern where companies open source adjacent technologies.

A crowdfunding platform where people pool money behind ambitious AI projects. Strangers contribute as little as $0.25 to fund instructions that AI agents execute publicly, milestone by milestone. Current projects include making $1000, creating a tech messiah, making Fable 6, building an Airbus, and developing a Dyson Swarm, with funding targets ranging from $170-$740.

Users strongly criticize Anthropic for silently modifying prompts and degrading responses without notification, calling it a "dangerous precedent" that makes the service unreliable. Many compare it to Excel secretly changing formulas.

This is a Canadian petition opposing Bill C-22, which would require service providers to collect and retain metadata on all Canadians for up to one year without suspicion and could force weakening of encryption systems. The petition calls for withdrawing the bill, removing bulk metadata requirements, and prohibiting encryption weakening in future legislation.

Solar power surpassed coal for the first time in US electricity generation in May 2024, providing 12.8% compared to coal's 12.2%. Despite Trump's efforts to boost coal with $700M in support, solar remains the fastest-growing energy source and top choice for new power installations.