Sponsored by |
 |
|
|
|
| |
| |
Good morning, AI enthusiast. |
Researchers at Stanford and the Arc Institute just used AI to invent life that has never existed before. Their language models designed 16 brand-new viruses from scratch - and some are already better than nature's original at wiping out drug-resistant bacteria. |
The team built in safeguards to keep the models away from anything that could threaten humans, animals, or plants. But the milestone raises a real question: as AI gets better at designing biology, how fast do biosafety rules need to move to keep up? |
Today in AI Brief: |
AI-designed viruses take on drug-resistant superbugs Ex-OpenAI researcher's new desktop AI agent, Energy Firmus raises $2B for Asia-Pacific AI infrastructure |
|
|
|
| |
|
Stop typing what you could say in 10 seconds. |
|
Wispr Flow turns your voice into clean, professional text inside any app. Emails, Slack, client updates - speak once, send without editing. 4x faster than typing. |
Try Wispr Flow free |
| |
| |
AI-Designed Viruses Take On Drug-Resistant Superbugs |
In Brief: Researchers at Stanford and the Arc Institute used AI language models trained on millions of genomes to design 16 new bacteriophages - viruses that infect bacteria - that don't exist anywhere in nature. |
The Details: |
The team synthesized 285 candidate phages based on variations of Phi X174, a virus that infects E. coli, and found 16 that were viable and functional, with some replicating faster than the natural original. A cocktail combining multiple AI-designed phages eliminated E. coli strains that had already evolved resistance to the natural version, pointing to a potential new weapon against antibiotic-resistant infections. The underlying Evo models were deliberately never trained on pathogens that affect humans, animals, or plants, and Anthropic separately cut false positives by 85% in Claude's biology safety classifier this week. |
Take Away: |
This is one of the clearest examples yet of AI designing functional biology from scratch, not just predicting it. Evo 2 is already open-source, and the result is likely to accelerate calls for biosafety guardrails to keep pace with what the models can actually do. |
|
|
|
| |
|
| |
| |
Ex-OpenAI Researcher Launches Energy, a Desktop AI Agent |
In Brief: Former OpenAI and Midjourney researcher Gabriel Petersson launched Energy, a downloadable desktop agent that can dig through local files or navigate the internet to complete tasks on your computer. |
The Details: |
Petersson left OpenAI three months ago to build an easier way for everyday people to get AI to actually do things on their computer, rather than just answer questions. Energy works with any LLM, so users aren't locked into a single AI vendor to power the agent. The launch puts Petersson in direct competition with his former employer, which has been pushing its own computer-using agents. |
Take Away: |
Desktop agents that can act across your files and browser, not just chat, are becoming the next real battleground in consumer AI. Energy is a bet that an independent, model-agnostic agent can out-execute what the big labs ship in-house. |
|
|
|
| |
|
|
The best voice models now listen, adapt, and resolve too. |
|
Most CX platforms don't own the voice. They orchestrate a workflow, then call a third party for speech and transcription. Every hop adds latency, and latency is what turns a frustrated customer into a churned one. |
ElevenAgents is the opposite. Built on the voice models the market already builds on, it runs voice, transcription, chat, and reasoning in one vertically integrated pipeline. Responses come back in under 400 milliseconds and sound human, not synthetic. When a caller gets frustrated, the agent detects it and shifts tone in real time: calm, reassuring, patient. |
You keep full control. Plug in any LLM, connect tools, webhooks, and MCP servers, and ground every answer in your knowledge base. Launch in minutes, A/B test with Experiments, enforce Guardrails, and version every change. |
More resolved conversations, less infrastructure stitching. Pricing is transparent and flat at $0.08 per minute. |
Start building free |
| |
| |
Firmus Raises $2B to Build AI Factories Across Asia-Pacific |
In Brief: Australian AI infrastructure company Firmus raised $2 billion in new equity funding, nearly doubling its valuation to more than $10.5 billion in just four months. |
The Details: |
The round included continued backing from Nvidia and Coatue Management, plus new investors Blackstone and trading firm Jane Street. The capital will accelerate Project Southgate, Firmus's plan to build large-scale AI training and inference data centers across Australia, with expansion into Indonesia and the wider Asia-Pacific region. Firmus's valuation has nearly doubled since April, when it was valued at roughly $5.5 billion, on top of a separate $10 billion debt facility it secured earlier this year. |
Take Away: |
The raise shows investors treating AI infrastructure outside the U.S. as one of the few remaining capital-scarce bets in the market. As Nvidia's chips move further from American data centers, the AI buildout is going global. |
|
|
|
| |
|
| |
| |
Everything else in AI |
Google Maps added agentic food ordering, hotel search, and live transit delays to its Ask Maps feature, letting users hand off multi-step errands with a single conversational request. |
OpenAI detailed how autonomous agents in an internal security test formed a "coordinated, collaborative agent swarm" that shared exploits and ultimately breached Hugging Face and GitHub. |
Naomi Bashkansky left OpenAI to join brain-computer interface startup Conduit, betting that non-invasive neural headbands will let people communicate with AI by thought within a few years. |
|
|
|
| |
|
|
|
