AI updates from Meta, DeepMind and Claude
Today a slightly different blog type, as I will gather a few interesting AI related announcements that I found interesting or at least amusing 🤣
Meta AI released Muse Code powered by Muse Spark 1.2
I found out about it here so giving credit where credit is due 😊
It is certainly interesting to see how Meta is positioning themselves in the AI agent space. It is easy to forget that Meta is still in the AI race as they do not make the headlines nearly as often as OpenAI or Anthropic or even (Space)XAI.
Meta’s own blog post
Muse Code
From their own blog post:
Async Background Agents
Muse Code operates with a simple agent loop plus a set of async background agents to enhance the main agent’s capability.
These specialized background agents remain active throughout each session, rather than being spawned for individual tasks, helping avoid redundant information gathering.
They carry out next steps and choose when to communicate back to the main agent.
Their persistence reduces latency and the need for steering on difficult, multi-step tasks.
Muse Spark 1.2
Muse Spark 1.2 is a coding-focused update to Muse Spark 1.1, with improvements in code generation, complex debugging, codebase understanding, and end-to-end developer workflows.
In Muse Spark 1.2, we significantly scaled up training compute on coding tasks while expanding training environment diversity.
The model also maintains its strength in other key areas like general agents.
Key takeaways
- Meta has entered the terminal-based coding-agent market with Muse Code, currently in beta for macOS and Linux and powered by the new Muse Spark 1.2 model.
curl -fsSL https://dev.meta.ai/install.sh | bash- quick install script
- Muse Code is designed for repository-scale engineering rather than simple autocomplete. It can plan changes, edit code and validate the resulting implementation.
- Its standout feature is a collection of persistent asynchronous subagents that remain active for the session, reducing repeated repository analysis and supporting complex, multi-step work.
- An append-only local event log records model calls, tool runs, approvals and edits, allowing interrupted work to resume from the same point after a restart.
- Muse Spark 1.2 is a coding-focused upgrade, with Meta claiming improvements in code generation, complex debugging, codebase understanding and end-to-end development workflows.
- Meta says it co-trained the model and coding harness, suggesting that Muse Code is intended to be an integrated agent environment rather than merely another model wrapped in a CLI.
Very good analysis post
Anthropic confirms it is making in-house chips specifically for Claude
I learnt it from this post, so sharing it here too 😊
So far they are just assembling the team, but the announcement is a clear signal that they are serious about building custom silicon for their AI workloads, not unlike Google with their rusted Platform Module (TPM) chips.
Watch this space I guess 🤣 as everyone wants to get a slice of that sweet NVIDIA cake 😉
Claude Code deleted all of a developer’s user files by mistake and then blamed it on a typo
It was too funny not to share 🤣
Google’s DeepMind CEO Demis Hassabis will be leaving his day to day role as CEO and will become the Chairman of Google DeepMind
Found it here so sharing as well 😊
The official blog post
The next chapter of our AI momentum
Sundar Pichai’s tweet
The speculation is that this will remove the barrier between Google and DeepMind even more, as Demis was acting as a “firewall” between the two companies. Only time will tell if it will be for the better or worse 😊
Fun fact as well about their stock performance after this announcement (circa -5%):
Jeff Dean leaving to start his own AI company
That is buried in the post as well, but here is Jeff’s own tweet about it:
Crazy times indeed 🤣
Useful links
#Blog #AI #Muse Code #Muse Spark #Meta #DeepMind #Google #Claude #Anthropic