
https://xitter.cf/t.co/36eFDHoV3o 🟠https://xitter.cf/t.co/rjCZMWrRMK #Art tools for creative economy
Joined July 2012
- Tweets26.2K
- Following378
- Followers10.5K
- Likes3.4K
Anthropic’s new research preview app lets Claude Opus 5.5 identify priorities, complete work independently, and check in before acting on your behalf, using your existing memory and connected tools.
Built with strong safety features, it’s available via waitlist for Pro and Max users in the US, with ongoing refinements based on early feedback.
claude.com/product/claude-du…
A Month of Thinking for $10
Until November 30, 2026
Google's new Gemini 4 Argon can think and write up to 1 million tokens in a single request
In human terms, that's roughly a month of focused thinking spent on one hard problem, like a tricky math proof, a deep research question or a long chain of reasoning.
At the launch price of 10 dollars per million output tokens, the whole "month" costs about ten dollars. From December 1, the price doubles to 20 dollars.
Argon is available now to paid API users and Google AI Ultra (PRO) subscribers.
mlearning.substack.com/p/how…
Google says Argon helped refactor Chrome, accelerate scientific software, and identify security vulnerabilities. Built for multi-step work across coding, finance, law, and science, it supports a unified one-million-token input-output limit.
Access is rolling out TODAY through developer tools and paid Gemini subscriptions (PRO, ULTRA)
mlearning.substack.com/p/how…
Claude just dropped monthly API credits for subscribers, and if you’re on Max 20x that’s an extra $200 every month
Claim it in Settings → Billing once your Console org is linked. New subscribers wait 7 days. In Claude Code you can spin up the prebuilt agents with /claude-api managed-agents-onboard.
$200 a month in API credit is a real upgrade
mlearning.substack.com/p/how…
Two former Google DeepMind researchers launched Beam, an open-weight system for coding, reasoning, and agents, with 501 billion total parameters and 23 billion active.
The full weights are due this month so a company can customize the model on its own data and its own machines.
The lab raised $4.6 billion and committed more than $7 billion to compute through 2029, including a $1 billion capacity deal and $150 million each month for a large training cluster.
It is expected to trail the most advanced closed systems in the US at the start.
mlearning.substack.com/p/ai-…

