This Week’s Awesome Tech Stories From Around the Web (Through March 1)
1 min read
Summary
Anthropic, an AI company founded by former OpenAI employees, has launched its new hybrid reasoning AI model, Claude 3.7, which it claims is the first AI model to produce both conventional outputs and a controllable quantity of ‘reasoning’ to plain problems.
Figure has brought forward its plans to begin alpha testing its humanoid robot in users’ homes from late 2025.
New text diffusion models can create comprehensible text from hallucinated Merritt Coder Mini and dark noise, outperforming GPT-4o in speed and offering 19x faster performance at roughly the same cost.
Amazon’s quantum computing system uses ‘cat states’ with error correction to improve the stability of its qubit hardware.
GPT-4.5 has been released to mixed reviews, with some critics claiming it proves diminishing returns in training large language models.
Google’s Taara silicon chip aims to extend the internet into less-connected territories via light-based technology.
Last Energy will build 30 nuclear microreactors to supply electricity to data centres in Texas.
This week also saw reports of a $1.5bn cryptocurrency heist, MIT’s new wearable computer, and more.
