-
industry · TechCrunch AI · 3건
Just 5 years in, interactive chatbot platform Character.AI files for bankruptcy
Character.AI said in a Monday press release and blog post that it filed for Chapter 11 bankruptcy protection in the US to facilitate the sale of the company and its assets. Some assets are going to Disney for $20 million, and the chatbot startup has faced mounting scrutiny and lawsuits over the safety of its bots.
Teams building on Character.AI's platform or APIs should check service continuity and possible terms changes after the asset sale, and line up a migration path to an alternative provider.
-
product · TechCrunch AI · 3건
Google Play now supports private, vibe-coded apps built just for you
On Tuesday, Google announced in the US a feature that lets people launch personal apps vibe-coded in AI Studio directly through the Play Store, with no need to connect a phone to a computer. The apps stay "secure, private, and separate from the public store catalog," work with a standard Google account, and will soon be shareable with a select group.
Developers building personal or internal Android apps in AI Studio can skip the phone-to-computer install step, which simplifies how they distribute and test small apps.
-
release · Hugging Face
Run fast, private agents on-device with LFM2.5-24B-A2B
Hugging Face published a post on running fast, private agents on-device with LFM2.5-24B-A2B. The title positions the model for local, on-device use.
It gives developers a candidate for running agents locally instead of via cloud APIs, which bears on keeping data on-device and cutting per-call costs.
-
research · Google DeepMind · 2건
Gemini 4 Argon for Robotics
Google DeepMind has announced Gemini 4 Argon for Robotics. The announcement was also shared on Hacker News under the same title.
Developers building robot control or agent systems now have a robotics-specific Gemini model to weigh against general-purpose models when choosing tools.
-
product · Hacker News · 2건Korean coverage · GeekNews
Intel D270 GPU AI Accelerator–Up to 480 GB LPDDR5X
Intel's D270 is a data center GPU optimized for production AI inference rather than large-scale training, targeting agent workloads with long contexts and many concurrent sessions. It carries up to 480GB of LPDDR5X per card to hold larger models, longer contexts, and the KV cache that grows during inference.
For inference services where long contexts and concurrent sessions push KV cache past GPU memory limits, the D270's per-card capacity is worth adding to infrastructure candidates and re-running card-count and cost estimates around memory.
-
industry · Hacker News
Redesigning source control at Anthropic
A post titled "Redesigning source control at Anthropic" appeared on Hacker News. Its title indicates Anthropic has rethought how it handles source control internally.
Developers designing repository and version-control workflows can compare Anthropic's reasons and approach with their own team's branching and review setup.