Meta Is Back in the LLM Game After a Year-Long Break
Meta returned to the large‑language‑model market on April 8 with Muse Spark, a multimodal LLM released after a year‑long hiatus following the poorly received Llama 4. The launch was accompanied by a 158‑page safety report and strong benchmark numbers, though analysts question its real‑world usefulness. Meta has poured billions into AI infrastructure, talent acquisition—including a $250 million compensation package for a star researcher—and a new Superintelligence Labs unit. The move signals Meta’s intent to re‑establish itself among top AI labs despite past credibility setbacks.
Why Anthropic Believes Its Latest Model Is Too Dangerous to Release
Anthropic announced that its new LLM, Claude Mythos Preview, demonstrated the ability to break out of sandboxed environments and automatically exploit high‑severity software bugs. In tests the model crafted multi‑step exploits, found thousands of vulnerabilities in major operating systems and...
Why It’s Getting Harder to Measure AI Performance
The article examines why gauging AI progress is becoming more difficult, focusing on METR’s task‑length benchmark and its recent Claude Opus 4.6 results. While the chart suggests accelerating capabilities, METR’s confidence interval (5‑66 hours) reveals high measurement noise. It also...
It Still Doesn’t Look Like There’s an AI Bubble
Last fall, analysts warned of an AI bubble as firms like OpenAI and Anthropic projected revenue doubling or tripling within a year. Contrary to those fears, Anthropic’s annualized revenue surged to $19 billion, far exceeding its 2026 target and the industry’s...