Google launches Project Suncatcher: TPUs orbit Earth for AI compute
OpenaI partners with AWS, Anthropic's plan to retire models, New Qwen reasoning model, Free Claude Code access, and Cogn
| Hey |
|
Welcome to AlphaSignal, the most read source of news by AI engineers and researchers. Every day, we identify and summarize the top 1% of news, papers, models, and repos, so you’re always up to date. Here’s today’s roundup: |
| Top News | |
|
|
|
| Google explores orbital TPU clusters that train models using constant sunlight instead of grid power | |
| 18,385 Likes | |
|
|
|
When Google says it’s sending TPUs to space, it’s not a metaphor. Project Suncatcher is a new research effort to build machine-learning infrastructure that runs entirely off the sun. The plan: clusters of satellites equipped with Trillium v6e TPUs, Google’s current cloud accelerators, orbiting Earth, linked by lasers, training models in constant sunlight. The Setup
The Problem
The Insight
The Breakthrough
The Impact
You can read the full technical breakdown in Google’s preprint “Towards a future space-based, highly scalable AI infrastructure system design.” |
|
|
| Deepnote, the AI Notebook, Goes Open Source | |
| Sponsored | |
|
|
|
Deepnote, now open source under Apache 2.0, lets you analyze data, build dashboards, and collaborate with AI, all in one place. Built as the successor to Jupyter, it stays fully compatible while adding a built-in AI agent, 100+ data integrations, and a human-readable .deepnote YAML format for clean version control. It allows you to: |
|
|
|
| partner with us |
| Signals | ||
|
|
||
|
||
|
||
|
||
|
||
|
|
| Top Repos | |
|
|
|
|
|
|
|
|
| Top Read | |
|
|
|
| Anthropic shows how to cut token use with MCP | |
|
|
|
Building agents that handle many tools often leads to high token use and latency. What it teaches: This blog by Anthropic explains how to use the MCP to run agent code outside the model context. It shows how to move loops, filters, and data parsing to an external sandbox that executes code directly and returns only key results. How to use it: Set up a secure code execution environment, organize your tool modules, and store state between runs. This helps your agent reuse logic and avoid repeated context loading. Why it matters: You build faster, cheaper, and more scalable agents that use fewer tokens without losing tool flexibility. |
|
|
|
|