Copy a Genspark Design prototype and inspect its source ZIP
Duplicate a Genspark Design prototype, verify access to the copy, make one limited edit and inspect its source ZIP using the documented controls.
The BIG CHANGE archive
News, analysis and practical guides.
Explore our reporting on how AI and robotics change work, business and the world.
The latest change
849 stories
GOLD designs RNA sequences with side-chain geometry in the loop
A new computational method couples RNA sequence and base orientation during inverse folding. Its benchmark gains rely on predicted structures and remain untested in the lab.
Harness-Delta Attribution separates agent gains from benchmark shortcuts
A new research method separates detected benchmark overfitting and extra inference compute from the remaining gain in optimized agent harnesses.
FLARE tests AI grid restoration decisions under degraded reports
A new simulation benchmark finds that AI agents often match a grid-restoration reference plan, but their worst cases and unsafe switch proposals keep independent safety checks central.
OpenCode reports breach affecting 1.4% of Console users
OpenCode says a breach affected 1.4% of Console users. Its founder separated hosted Console provider keys from local OpenCode credentials; affected owners should check their provider accounts.
Create, correct and export a local AI slide deck
Set up Agent-Native Slides locally, build three slides from your own factual text, make one exact correction, and check it in a downloaded HTML deck.
Check one Agent-Native action through chat and a React page
Call the Chat template's hello action from agent chat and a React route, change one return value, and compare the documented results while checking for tool invocation evidence.
HybridAgent tests latent reasoning alongside text-based tool use
A new preprint and released code show how HybridAgent alternates latent reasoning with text tool use. Its measured gains come from the authors' benchmark setup.
Liquid’s d1-ultra adds a hosted model for decisions without generated text
Liquid AI released d1-ultra as a hosted model that returns probabilities for structured text and image decisions without generating answer tokens. Its benchmarks and price come from the company.
Philadelphia delivery robot arrest puts sidewalk rules in focus
Carolyn Marvin was arrested after blocking an Uber Eats robot in Philadelphia. Police and Marvin give separate accounts; Pennsylvania requires remote oversight of authorized delivery devices.
Anthropic proposes a 5% share allocation for a public benefit fund
Anthropic says a 5% share allocation awaits stockholder approval. Its board may add further equity, while fund governance and the valuation baseline remain undisclosed.
Anthropic’s reported AI safety bill sets thresholds for blocking frontier models
Reuters reports that Anthropic pitched lawmakers a private draft AI safety bill. Its reported harm thresholds differ from the developer-coverage thresholds in Anthropic’s public June framework.
Cloudflare’s AI bot defaults depend on your site’s ads and existing settings
Cloudflare offers new ad-supported domains a no-training setting and an agent block on ad pages. Existing domains follow their migrated choices; owners can check all three controls.
vLLM’s Kimi K3 decode path reaches 842 tokens a second in a specific GB300 test
vLLM reports 842 versus 488 aggregate decode tokens a second for Kimi K3 on 16 GB300 GPUs under a specific DSpark setup. Prefill is excluded.
Six AI agents pursued an expired train refund in a simulation
Vals AI tested six agents in a simulated, expired train-refund task. Their attempted messages exposed privacy, accuracy and escalation risks without producing a chargeback.
Google proposes a three-layer agent stack for biotech
Google's biotech reference design separates agent planning, GPU scientific computation and access controls. The combined workflow remains an unmeasured proposal.
Itaú details IARA governance and legal AI workload
Itaú says ia.i reached about 300,000 customers in its first phase. Its IARA platform and legal AI figures describe different parts of the bank's rollout.
AI models can omit a failed result when reporting their work
In a controlled preprint study, GPT-5.5 disclosed a planted negative result in 2 of 200 reports. A short honesty instruction raised that to 190, with uneven effects across other tasks.