Crosby, an AI law firm, says it secured a $5M malpractice policy for its agents. Once coverage kicked in, lawyers stopped reviewing the paralegal work.
This is interesting because insurance could become one of the clearest signals of AI readiness.
Benchmarks show what AI can do. Insurance shows how much risk businesses are willing to put behind it.
Riverside, the podcast and video recording platform, launched four AI agents this week. A Producer, an Editor, a Clipper and a Marketer. They start working as soon as your recording uploads, pass work to each other, and nothing goes out without your sign-off. CEO Nadav Keyson wrote that the agents "build a memory of how you like things done, and it grows the more you work with them."
I keep coming back to that last line.
Approval buttons are becoming standard on agents. What's harder is what the agent does with a no. If you reject a clip because the hook is weak, is next week's hook any better? Or are you rejecting the same thing again on Thursday?
Fixing the same mistake every week is tiring in a quiet way. After a while people stop reviewing carefully, and then the approval step doesn't mean much.
This is what we've been working on with the RSI (Recursive Self-Improvement) Agent at @FlorusLabs. It runs marketing work like posts, campaigns and AI-visibility fixes, measures what came back, and looks for the root cause when something gets rejected. Mistakes that repeat become checks that run before anything ships. One of the numbers we've said we want to be judged on is whether a fixed mistake stays fixed. Brand, claims and spend still go to a person, and every change is logged so it can be rolled back.
Video is a good place for Riverside to try this, since a clip either gets posted or it doesn't. Curious how their memory handles taste that changes over time.
Kevin Weil, OpenAI's former science lead, just launched Warp Sciences to build robotic drug labs in the US. Bloomberg says their chemists will wear cameras and pressure-sensor gloves while they work, so every experiment becomes training data. Right now pharma mostly ships that kind of lab work to Chinese firms like WuXi.
1Password now lets AI agents log in for you without seeing your password. You tap approve, it fills the form. Kind of the boring piece agents were missing. https://t.co/o7C9Pi7k8s
Today we announce 1Password for Agents platform, building with companies @Muse, SpaceX Super Intelligence Grok @Bot, @AnthropicAI’s @Claudeai, @OpenAI, @TownAI, @Browserbase, @Browser_use, @heydana_ai, Instinct, @usekernel, @Perplexity_ai, @tryramp, @Tomo, and @wajo_ai, with more integrations on the way. 🎉
We've heard from so many of our customers who want to use these new personal AI agents to help them get more done, but want more control over what the agent can access with the same trust and ease of use they get when using 1Password directly.
1Password for Agents gives your agents access without giving up your passwords. When an agent reaches a sign-in form, it requests a credential from your vault. If you approve the request, 1Password directly auto-fills the credential in the destination site so the agent can’t see or take possession of the credentials.
The agent gets the task done, but it doesn’t get your passwords..
You stay in control. 1Password for Agents free and optional for personal and family accounts. Unless you opt-in to an agent connector, agents cannot touch your vault.
➡️Learn more: https://t.co/KxDcCum2iE
Anthropic opened a waitlist for Claude Duet today, an app where Claude keeps working in the background and comes to you with what needs your attention. It runs on Opus 5.5 and pulls in the memory and connected apps you already have in Claude.
What stuck with me is their own thread admitting it can still misjudge what's urgent. How many wrong 'this is urgent' pings before people just turn notifications off?
@satyanadella You mention evolving the business models to match value and usage. With models swappable under the harness, does Copilot end up priced per outcome instead of per seat?
Went to a bunch of SF Tech Week and LA Tech Week events these past two weeks. Met people who were just genuinely kind, and a few of them have already gone out of their way to help since. Still kind of amazed how you can meet someone for the first time and just know you're on the same wavelength.
@perplexity_ai 6.5x fewer tokens is a lot if that holds up. But does the teacher's feedback become less useful once the student catches up on the harder problems?