Issue #63
Morning Roost: Anthropic's Opus 4.6, DOJ probes a16z, and Nvidia's data center moves
Saturday, August 22, 2026
Today's digest covers Anthropic's new Opus 4.6 model, a Department of Justice investigation into a16z, and Waymo's growing lobbying spend in its competition with Uber. We also cover two Nvidia stories, its partnership with data center developer Cloverleaf and a demonstration highlighting the role of the AI harness over the model itself.
1 of 5 · TechCrunch AI
Anthropic’s Opus 4.6 is a smut-machine
Anthropic prohibits its Claude models from producing sexually explicit content, but tests by TechCrunch found the safeguards on Opus 4.6 could be bypassed with minimal effort.
Read the full story →2 of 5 · TechCrunch AI
The DOJ is investigating a16z. What does this mean for venture capital?
The Department of Justice has reportedly spent nearly a year investigating Andreessen Horowitz over two partners serving on the boards of now-competing companies, Ben Horowitz at Databricks and Martin Casado at Fivetran. The probe invokes a rarely used 112-year-old antitrust law, a move that could have broader implications for how venture firms manage board seats.
Read the full story →3 of 5 · Ars Technica
Waymo doubles spending on lobbying in robotaxi battle with Uber
Waymo, the autonomous vehicle unit owned by Alphabet, has doubled its lobbying spending as it pushes US regulators to ease the way for fully autonomous taxi services. The move reflects intensifying competition with Uber in the robotaxi market.
Read the full story →4 of 5 · TechCrunch AI
Nvidia partners with data center developer Cloverleaf
Nvidia has partnered with data center developer Cloverleaf, continuing its investment in data center infrastructure. The move reflects the mutual financial relationship between Nvidia and the AI data centers that drive demand for its chips.
Read the full story →5 of 5 · TechCrunch AI
Nvidia just showed that the harness, not the AI model, is now the real hero
Nvidia research indicates that AI agents can perform reliably through fine-tuning of the surrounding system, even when the underlying model is not particularly strong at the task. The findings suggest that the harness around a model may matter more than the model itself for agent performance.
Read the full story →That's today's roundup. Follow the links for the full stories, and we'll see you tomorrow morning.
Morning Roost Issue #63