I've been giving my agent it's own browser, hosted in docker, but accessible via web VNC for setup and observability. You login to your stuff, get things setup the way to like them, then let the agent use an MCP connection to control the browser.
Always on, easy to debug.
github.com/mdp/remote-browse…
This is an insane way to read this article. Agents are acting on the behalf of users who authorized them. Maybe the banks aren't comfortable with this, and that's fair, but this reads like a FB post from an older relative.
Also, that emdash :/
This is insane. We cannot allow AI agents to go into your bank account and act independently without your authorization. But that is what’s happening right now.
Left unchecked, AI could endanger the entire global financial system — including your bank account.
Pause AI NOW!
Readers added context they thought people might want to know
Per Axios, detected AI agents access bank accounts with the customer’s permission. This violates most banks’ terms requiring human logins, which is why banks call it “unauthorized.”
Source: Axios (Oct. 22, 2026) — axios.com/2026/10/22/ai-…
The jailbreaks keep coming. For stuff like Kindles, it’s worth taking advantage of this window of opportunity while it exists. Lock in that freedom while the AI is ahead of the firmware.
pepsipu.com/blog/2026-10-swi…
This was my first thought, you now have an easy way to MITM everthing shitty in your house. I think we are in for a fun time when people start posting what they're finding.
I always use the last 4 of my social for my ecc keys. Last 6 if it’s work related and requires a bit more security.
This whole thread on the Deepseek infra release is a great read, especially if you're curious about GPU economics - "the service throughput of a single cluster is 155 trillion tokens/day, and the overall revenue is $38 per card per hour (based on off-peak prices)"
news.ycombinator.com/item?id…
Mark Percival retweeted
We’re betting our $5.5M seed round that the next leap in AI coding won’t come from OpenAI or Anthropic.
It’ll come from a new feedback loop that learns from you.
Introducing Beacon: open-source, self-improving memory for coding agents.
Your sessions capture how you debug and solve problems. Today, that context gets lost and you have to re-explain things every time you switch tools.
Beacon continuously distills your sessions into reusable skills across Claude Code, Cursor, Codex, and 20+ harnesses, so your next agent can learn from the last.
Beacon is free and OSS 🌟 github.com/Asymptote-Labs/ag…
Why should the frontier labs be the only ones winning with your sessions?
You already did the work. Now let your coding agents learn from you.
I'll be at Cloudflare Connect this week, who else is going to be there? Shoot me a DM if you want to grab coffee / meetup.
Do yourself a favor and refactor your Agents.md files - it's not the same world today. Give your agent the tools and let them cook. Very little direction is needed.
call me crazy but i actually think if you’ve been vibe coding for the last 12 months you’re at a disadvantage
your expectations of what models like opus 5.5 can do are probably too low and your workflow are probably too complex for their own good
you can literally just tell it what you want and it’ll do it. literally that’s it
Mark Percival retweeted
all the companies that have lost goodwill are going to get commoditized by hordes of angry users out of spite
companies that do have goodwill will be just fine, they can even opensource their moat and people will be happy to give them money
case in point, adobe versus tailscale
If you're into retro gaming, there's a real renaissance happening right now with "decomps" and "recomps" - It's actually kind of magical to see this happening to all my favorite old games.
heldgames.com/guides/retro-d…
And if you're looking for an easy way to get started, you can use this open source project to try out the games - github.com/tgeorgiadis/quive…
Look at this, Contra decompiled! github.com/vermiceli/nes-con…
Literally, my childhood, decompiled.
I don't like having the agent using my local browser (and I don't want it local) - So I put it in a docker container then set it up how I like it (logged into what I care about) via a web based VNC.
github.com/mdp/remote-browse…
















