LastWeekIn.Tech

A mixed news day featured sharper scrutiny of AI economics and infrastructure friction alongside fresh research on self-improving multi-agent systems and surveillance-tech governance battles.

1. 1

   Number 1. Industry & Economics Hacker News

   Research Firm Calls Anthropic's IPO 'Unprecedented Test of Investor Gullibity'

   Independent research firm New Constructs called Anthropic's upcoming IPO the 'most ridiculous IPO of 2026,' valuing the company at roughly $150 billion versus a reported target near $2 trillion. The firm argues Anthropic would need to generate more than double Nvidia's trailing-year net income to justify the sought valuation.

   Why it matters The critique spotlights a widening gap between AI-sector valuations and underlying financial fundamentals just as Anthropic prepares a high-profile Nasdaq listing. If investors push back on the pricing, it could ripple through comparable AI valuations across the industry and chill enthusiasm for upcoming AI IPOs more broadly.

   What to watch Watch for Anthropic's S-1 filing details and whether institutional investors signal pricing concerns ahead of any listing.

2. 2

   Number 2. Infrastructure & Chips Hacker News

   Google told to halt work on datacentres in Finland over environmental concerns

   Finnish authorities ordered a halt to construction on two Google-linked datacenters in Muhos and Kajaani, part of a €13 billion northern Finland buildout, after the company representing Google failed to complete required environmental impact assessments. Reports note 330 hectares of forest had already been felled at the Muhos site without the EIA in place.

   Why it matters The stop-work order illustrates mounting regulatory friction in Europe around the land, energy, and environmental footprint of the AI compute buildout, and sets a precedent for enforcing environmental review before hyperscale construction proceeds. Delays here could slow planned AI capacity expansion in the Nordics, a region favored for cheap, green power.

   What to watch Watch whether Google completes the EIA process quickly or faces further delays, and whether other EU jurisdictions tighten similar pre-construction environmental requirements for datacenters.

3. 3

   Number 3. Capabilities Hacker News

   Recursive self-improvement through collective intelligence

   Researchers introduced MASS, a method where AI systems act as both optimizer and evaluator to repeatedly refine multi-agent workflows, then distill the resulting collective intelligence as a self-supervision signal. The approach reportedly improves score-per-output-token over base models on research benchmarks without relying on human-labeled data.

   Why it matters This is a concrete technical step toward recursive self-improvement architectures, where AI systems optimize their own agentic workflows rather than depending on human curation. It feeds directly into ongoing safety debates about self-improving systems, since methods like this could accelerate capability gains while reducing human oversight of the optimization loop.

   What to watch Watch for independent replication of the benchmark gains and whether safety researchers flag risks from using AI as both evaluator and optimizee at scale.

4. 4

   Number 4. Policy & Regulation Hacker News

   A Chicago-area college told Flock to remove its cameras. The company refused

   Elgin Community College in Illinois asked Flock Safety to remove six AI-powered license-plate-reading cameras after concerns emerged that captured data was being shared beyond its original stated purpose. Flock Safety refused to remove the cameras, citing its contract with the college.

   Why it matters The standoff exposes a growing governance gap around AI surveillance vendors: institutions that deploy AI-driven monitoring tools may find themselves locked into contracts even after discovering unexpected data-sharing practices. It raises questions about consent, data control, and accountability as AI-powered license-plate and facial tracking systems proliferate across public institutions.

   What to watch Watch for legal action or legislative response addressing contract termination rights and data-sharing transparency requirements for AI surveillance vendors.

5. 5

   Number 5. Industry & Economics Hacker News

   Attack of the Vibe Coded Photoshop Clones

   A tracking report documents roughly a dozen 'vibe coded' AI-generated clones of Adobe's Photoshop, plus dozens of forks, with similar cloning efforts now extending to Microsoft, Avid, Autodesk, and game-development software. The piece frames this as automated or semi-automated replication aimed at siphoning revenue from established software vendors.

   Why it matters This underscores how generative AI coding tools are lowering the barrier to rapidly replicate commercial software functionality, threatening established licensing and subscription revenue models across the creative and productivity software industry. It foreshadows likely IP disputes and platform-level policy responses as AI-assisted cloning scales beyond a niche phenomenon.

   What to watch Watch for cease-and-desist or litigation activity from affected software vendors and any marketplace/platform rules targeting AI-cloned commercial app distribution.