← Back to archive

Wednesday, August 26, 2026

DeepMind spinout just embarrassed OpenAI & Anthropic

Inherent, a DeepMind spinout, just dropped a tiny AI agent that beats both OpenAI and Anthropic at replicating scientific research (bold move for a new player), while Nvidia eyes a massive $30B+ investment in Perplexity with a tech licensing sweetener. Meanwhile, one startup is training AI on actual living human skin tissue for a month at a time to discover cosmetic ingredients (yikes). So here's the thing: when a scrappy startup outperforms the giants at research replication, are we finally seeing the democratization of frontier AI, or just better benchmarking?

Top Stories

1
Inherent, founded by DeepMind alumni, says its AI 'teammate' just outperformed Anthropic and OpenAI at replicating research

TechCrunch

DeepMind alumni-founded Inherent claims its AI agent Faraday outperformed much larger models from Anthropic and OpenAI at replicating scientific research, using a fraction of the computing power by employing reinforcement learning to develop "research taste." The startup, which raised $50 million in seed funding, aims to build AI systems capable of discovering new scientific knowledge.

agentsdeepmindfundingreinforcement-learning
2
SoftBank Plans Record ¥1 Trillion Retail Bond Offering in Japan

Bloomberg

SoftBank is preparing a record-breaking ¥1 trillion retail bond offering in Japan, the largest such sale in the country's history, as it seeks capital for AI and technology investments.

softbankfundingjapaninvestment
3
Nvidia Discusses Perplexity Investment, $30 Billion-Plus Valuation Considered, Tech Licensing Deal

The Information

Nvidia is discussing investing in AI search startup Perplexity at a $30+ billion valuation, more than triple its previous $9 billion valuation, potentially including both equity investment and a technology licensing deal.

nvidiaperplexityfundingvaluation
4
A startup trains AI on living human skin tissue

TechCrunch

Outer Biosciences keeps living human skin tissue viable for over a month and uses AI to rapidly discover new cosmetic ingredients, generating candidates every six weeks compared to the industry's 18-month timeline. The startup has raised $23 million and plans to license its discoveries rather than build consumer products directly.

biotechaidrug-discoveryfunding
5
Thinking Machines offers $50,000 in grants for open-weight AI safety research

Thinking Machines launched a $50,000 grant program to fund safety research on open-weight AI models, focusing on tamper-resistant safeguards, hazardous data filtering, differential defensive capabilities, and measuring worst-case risks after model release.

open-sourceai-safetygrantsopen-weight

Keep Reading

Industry Voices

Enjoyed this issue?

Get daily AI intel delivered to your inbox. No fluff, just the stories that matter.