For business owners
Faster, cheaper AI for agents, and a warning for your servers
November 10, 2026 in summary: Anthropic's Claude Fable 5.2 runs up to 30 percent faster at 20 percent lower prices, Meta's Muse Ember is built for the repetitive work agents do all day, and an actively exploited OpenSSH flaw is a reminder to patch the infrastructure underneath it all.
- Daily brief
- AI news
- Automation
November 10 was a day of two halves. The first half was good news for anyone who wants AI to do real work: two new models arrived that are faster, cheaper, and built specifically for the repetitive jobs agents do all day. The second half was a warning: a security flaw in the software that runs much of the internet is being actively exploited, and the fix is already out. Put together, the day tells a familiar story. The tools keep getting better and cheaper, and the infrastructure underneath them still needs someone paying attention.
The agent economy is getting cheaper
The headline launch came from Anthropic. Its new Claude Fable 5.2 produces outputs up to 30 percent faster than version 5.1, Fable is now the top scorer on the Artificial Analysis index, and prices are down 20 percent, according to MacRumors. That combination matters more than any single number. Speed and price are the two things that decide whether an AI agent can run all day doing useful work instead of being a demo you switch on for ten minutes. A model that answers faster and costs less per call changes the math on automating the small, repetitive tasks that fill a business day: triaging email, drafting replies, sorting documents, following up with customers. When the per-task cost drops by a fifth and the wait shrinks by nearly a third, work that was barely worth automating last month starts to make sense now.
Meta made the same bet from a different direction. Its new Muse Ember is the company's smallest and cheapest model yet, built for exactly the kind of work agents do all day: tool calls, coding subtasks, and extraction, at 10 cents per million input tokens, according to TechCrunch. Ember is not trying to be the smartest model in the room. It is trying to be the cheapest one that is good enough for the boring jobs, and that is the point. An agent that checks a system, files a form, or pulls a number out of a document does not need a flagship model. It needs a model that costs almost nothing per call and never gets tired. Meta pricing that work at a tenth of a dollar per million tokens is a signal that the industry now sees agent execution as a volume business, where the winner is whoever can do the most small jobs for the least money.
Read the two launches together and the direction is clear. Yesterday the story was about power and scale; today it is about cost and speed at the point of use. The frontier labs are no longer competing only on who can build the biggest model. They are competing on who can make an agent affordable enough to leave running.
The plumbing still needs patching
The other story of the day is a reminder that none of this runs on magic. Hackers are actively exploiting a flaw in OpenSSH, nicknamed SSHole, that lets them log into passwordless accounts on Alpine Linux and other musl-based systems, according to BleepingComputer. The fix is OpenSSH 10.6p2, and if you run affected systems, the advice is simple: update now.
This is the unglamorous side of the AI boom. Every agent, every automation, and every cloud service sits on top of infrastructure like OpenSSH, and a flaw there can undo a lot of careful work elsewhere. Alpine Linux is popular precisely because it is small and efficient, which is why it shows up in containers and small servers everywhere, including the ones small businesses run without thinking about them. A passwordless account that an attacker can log into is a door left open, and the exploitation is already happening, not theoretical.
What this means for a small business
Three practical takeaways. First, re-price your automation ideas. If you looked at an AI agent for a repetitive job a few months ago and the cost did not work, look again. Fable 5.2 at 20 percent less and Ember at 10 cents per million input tokens mean the same job now costs meaningfully less to run all day. Second, match the model to the job. The expensive flagship model is for the hard thinking; the small, cheap model is for the thousand small calls an agent makes while doing its work. Using the right one for each keeps the bill down without giving up quality. Third, patch your servers. If your business runs anything on Linux, check whether OpenSSH needs updating to 10.6p2, especially if you use Alpine-based containers. The new models will save you money, but only if the systems they run on are not compromised.