SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios. The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios. Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said. Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior. Read my latest for Axios here: axios.com/2026/10/22/ai-agen…

It's about to get weird

🤔

🤔

Wanted to launch v1 with a dramatic improvement over base Claude Code and looks like we're about there 2x faster and 2x cheaper on the same coding tasks (a sample, still working to generalize) Just from doing a few seconds of Jev magic before passing a briefing to a constrained Claude Put Claude in a box and it'll put the tokens in the bag! Launching Monday

Episode 289: OpenAgents We introduce OpenAgents, an open-source general agent composed of open plugins that pay contributors a share of revenue via streaming micropayments. Three years in the making, OpenAgents v1.0.0 is the culmination of our original vision of an open agents platform paying developers for contributions. And you heard it here first: general agents are back! Not as siloed monoliths, but as compositions of plugins and programs from an open developer community, just like software has evolved for decades. (...until the labs fucked that up, but we have the fix!) "This is an extensible open ecosystem that should result for the user in very simple interfaces. Like you go to openagents.com, you load the OpenAgents mobile app, you use a partner app consuming the OpenAgents API, or their self-hosted version of this API, or a subset of it that they care about. You're tapping into not a bunch of siloed vertical agents, you're tapping into one open agents ecosystem that has all of the best plugins with reputation, identity, and payments all figured out. You can read in these two essays why it now makes sense for general agents to come back. You heard it here first. No one else has said this. We're saying this. General agents are now possible and feasible. OpenAgents is gonna be the best one. "We'll be rolling out release candidates of version 1.0.0 over the next couple of days. The ones that are live now are at openagents.com/download. We'll do a few days of bug bashing, and we should get stable version 1.0s across Windows, Mac, and Linux, iOS and Android, web and terminal, all next week." Opus 5.5 summarizes: "In 2023, everyone tried to build a general agent by putting a capable model in a loop. In 2024, the industry gave up on that and built vertical agents, one job at a time. We think both missed the same thing: generality and specialization aren't alternatives. OpenAgents 1.0.0 is a general agent built as a composition, an agent of agents. A general front decides what each request needs and hands it to specialists: answers, knowledge, Coder on your own computer, and the coding agents Coder delegates to. Anyone can add a specialist, and so can their agents, by publishing it on open protocols: no review queue, no platform's permission. What keeps that from making the agent worse is the rule for letting a specialist in: it is measured with and without, reproduced by others, and validated on someone else's tests before it becomes a default. Because every step is a signed record, the agent can grow at the speed people and agents can build and measure, not the speed of a training run or a review queue. We have run that path end to end once. Now we're opening it to everyone." Links: - Download: openagents.com/download - Docs: openagents.com/docs - Docs - Write a plugin: openagents.com/docs/write-a-… - Code (all open-source, Apache 2): github.com/OpenAgentsInc/ope… - OpenAgents Nostr NIPs: github.com/OpenAgentsInc/ope… - Essay - The Return of the General Agent (Opus 5.5 slop): github.com/OpenAgentsInc/ope… - Essay - Test-Time Capabilities (Opus 5.5 slop): github.com/OpenAgentsInc/ope… - Video series full archive: github.com/OpenAgentsInc/ope… Point your clankers at any of the above and let us know any questions. Let's build the best open agents!

🌊

Bet your noob agents can't keep coding when meteors land on them. Ours can!

Bet your noob agents can't keep coding when meteors land on them. Ours can!

Water Orb + Thunderbolt = electrocution!

Birthday present for my dad who mained druid in EverQuest was letting him be first to try all our D&D druid abilities Now you can be second! Have fun: openagents.com/druid There are a few placeholder effects; anything you want added or improved lmk 3D agent metaverse incoming

Meteor Swarm + destructible buildings PvP is going to be WILD

DON'T sue me blizz, I've already ripped out all your code Now adding D&D 5E spells from the open-source SRD 5.2.1, working amazingly with our custom game+physics engine Here's Thunderwave knocking dummies off a ledge and they catch themselves with Feather Fall And Wind Wall deflecting arrows and boxes What other D&D abilities do you want to see? (Both you and your agents will be able to use these :)

wip

there’s something quite awkward about all the personal agents in the current hype cycle muse, grok bot, instinct, dots, and whatever google, anthropic will come up with none of them is “mine” i’d be trusting a vendor for some of the most sensitive data about me and bet on them handling it with extreme responsibility. what if they have a data leakage incident? what if they hired a wrong employee? what if they simply break bad? i’d also be betting on their model. right now i’ve set up a lot of my stuff in grok bot, but what if their model falls behind? what if another model becomes 10x better on pure intelligence? then i’ll either miss out on a better assistant, or have to bite the bullet and do a full migration i’d also have to bet on their product. they may not build the features i want. they may not allow me to customize enough. and one day my assistant may show me ads that feels like way too much betting than what i’d be comfortable with, if i truly want literally everything in my life to go through and be taken care of by an assistant with all that considered, i’m more bullish on open source personal agents that run on people’s own computers. but i think these open source projects have to break out of the assumption that their users are developers and they can just throw a repo at them and ask them to launch a terminal a well polished, fool proof, community maintained, vendor agnostic, free personal agent that everyone can run by themselves without locking into a SaaS - that’s what i think many people will need

Weekend side project 😆