techl8

you're late
Friday, 18 September 2026
AITechCrunchJust broke

PrismML releases Bonsai 2 27B, which compresses Alibaba's Qwen3.8 27B

PrismML releases Bonsai 2 27B, which compresses Alibaba's Qwen3.8 27B to 5.9 GB, small enough for smartphones, while retaining 98.2% of Qwen's benchmark scores

TL;DRNew compression technique shrinks a 27B AI model to 5.9GB while keeping 98.2% performance.

Why it matters: Enables powerful AI models to run locally on phones without cloud dependency or connectivity needs.

Latest

200 stories
4h

Anthropic says Claude "leads" 26% of its AI R&D work, up from 1%

Claude now leads a quarter of Anthropic's AI research work, up from 1% months ago.

3h

IQM Expands to South America with On-Premises QPU Sale to Brazil’s

2

IQM selling quantum computer to Brazilian research institute; first private QPU in region.

51m

Alice & Bob Partners with CEA to Integrate Cat Qubit Software into HPC

Quantum startup integrates cat qubit software into supercomputers for hybrid computing by 2030.

1h

U.S. Lawmaker Ro Khanna Urges Chinese Labs to Help Pace Frontier AI

U.S. congressman asks Chinese AI firms to slow development pace via international agreement.

1h

Bug Hunters Used Claude to Hack OpenAI

Researchers used Claude to breach OpenAI's systems and access internal code repositories.

1h

A federal judge denies OpenAI's bid

Judge blocks OpenAI from accessing Apple-SpaceXAI settlement details in antitrust dispute.

1h

Waymo Plans to Offer Paid Robotaxi Rides in Singapore

Waymo is launching paid robotaxi service in Singapore by 2028.

1h

Waymo says Singapore will be its next international robotaxi city

Waymo is launching self-driving taxis in Singapore by 2028.

2h

Indian Project Financier’s Data Center Loans Top $1.2 Billion

India's state financier is backing data centers with $1.2B in loans to boost AI infrastructure.

2h

What was Hard Fork?

Don't be late twice

Pick your categories and frequency. We'll handle the rest.

AI
Tech
Quantum
Robotics