Jürgen Leschner @jldec

Helping build a better Internet for agents and humans @cloudflare 🧡 jldec.me Joined May 2008

Tweets

3K
Followers

399
Following

2K
Likes

2K

Basil🧡 @LinkofSunshine

6 days ago

Idk why this comic gets me every time lol

187 2K 71K 2.5M 6K

View Details

sam @samgoodwin89

a day ago

How to embed an Alchemy-infused Agent into a Durable Object.

1 1 46 4K 26

0 0 1 60 0

View Details

Astro @astrodotbuild

a day ago

Introducing Sätteri — High-performance Markdown & MDX processing for the JavaScript ecosystem. satteri.bruits.org

25 78 935 42K 461

View Details

I actually think this is the wrong approach to agent authorization. Here's why: If you have to explicitly configure each agent's permissions, you've lost. Because you're only going to have patience to configure so many agent permissions. So in this route you can only have a certain relatively small number of agents before configuration fatigue prevents you from making more. I don't think that's what we want. I don't think that's what's good for AI safety. What we want is an enormous number of very fine-grained agents. Each task is a new agent. And each task has exactly the permissions needed for that task, no more, no less. There's really only one known way to make that manageable: Capability-based security. The basic idea is, when you give the agent a task, you naturally give it the capabilities it needs to perform that task. Like say you want an agent to review a Google Doc. Today, with a lot of AI assistants "Hey go review the document titled Foo Spec". The agent has permissions to all your docs, so it goes and finds the right one and opens it. That's wrong. You should say "Hey, go review this document: " And then here's the key part: The harness should see that you're pasting a URL, and should infer that you want to give the agent access to that document. Only that document. No other document. Importantly, you didn't really have to do anything unusual to configure this. You just pasted the URL of the thing you wanted the agent to access. Which you probably would have done anyway. Sure, it's not always that easy. Maybe you commonly run agents that need access to 10 different things, and it's tedious to paste those 10 URLs every time. So you create some sort of a bundle that you give them. And of course, the agent should be able to ask for extra things it needs. But we need to get away from this idea that the agent always starts out with access to everything, even when it doesn't need most of it. Also, I tend to think all agent authority has to derive from a human -- contrary to what is argued here. Every "autonomous agent" has to report to someone, and uses a subset of that person's authority. This is needed for accountability -- because agents are not accountable. If you see "Claude deleted the database", what are you supposed to do about that? You need to see "Claude acting on behalf of Bob deleted the database". To be clear, I totally agree that it's problematic when Alice configures an agent with her own credentials and then Bob tells the agent to do something with those credentials. Then you'll see "Claude acting on behalf of Alice", but actually Claude was acting on behalf of Bob. The answer is that Bob should not be able to command Alice's agent. Bob has his own agent, which may have all the same context, but operates with Bob's credentials. But if Alice sets up an agent for her team, Bob maybe doesn't want to spend time configuring his own version of it with all the same credentials, that's tedious. This has to be automated. I think capabilities make this easier. Alice gave a set of capabilities to her agent. The harness should be able to look at that list, and recreate the same list using Bob's credentials, without Bob having to do much except click "OK". I realize there's a lot of hand-waving here -- this is a complicated topic. I'll drop some code next week. If you're attending AI Engineer in SF, come to my talk on Tuesday, where I will also only be able to scratch the surface in 20 minutes... claude.com/blog/agent-ide…

84 56 675 83K 759

View Details

Cloudflare @Cloudflare

7 days ago

Today we're rolling out Temporary Accounts on Cloudflare Workers. Any agent can now run wrangler deploy — temporary and get a live Worker in seconds. cfl.re/4ekY0yV

30 188 1K 771K 932

View Details

Jürgen Leschner @jldec

a week ago

Venus

0 0 3 24 0

View Details

sunil pai @threepointone

a week ago

might pivot to this

Adam Wathan @adamwathan

a week ago

Feels like we're all just building "things for building other things" and not a lot of "things" anymore.

213 181 3K 249K 200

6 2 85 10K 9

View Details

korinne @korinne_dev

2 months ago

Had such an interesting conversation about this topic today with @amruthagujjar - and she immediately whipped up this beautiful article. "Static files were originally abandoned because personalization required compute at read time. Agents invert that equation: personalization now happens at write time. The result is a static file that is as personalized as any dynamically rendered page, but trivially cacheable, globally distributable, and entirely free of server-side compute at delivery."

Amrutha Gujjar @amruthagujjar

2 months ago

x.com/i/article/2051…

0 0 2 546 1

0 1 2 440 1

View Details

Brendan Irvine-Broque @irvinebroque

2 weeks ago

rita kozlov 🐀 @ritakozlov

2 weeks ago

the biggest voice / video chat app you know runs on @Cloudflare compute discord.com/blog/how-we-mo…

10 31 205 21K 63

5 7 64 8K 14

View Details

Anni Wang @aninibread

2 weeks ago

Working on @Cloudflare R2 is probably some of the most fun I’ve had. If you want to work with a world-class eng team and exponentially growing product do apply! If you have any questions about this product role and team, feel free to reach out to myself or @fireblaze_15!

rita kozlov 🐀 @ritakozlov

2 weeks ago

i'm hiring a pm for @Cloudflare r2 if you know anyone who would be a good fit lmk

7 15 133 50K 13

1 1 15 844 1

View Details

kyle @kylecarbs

2 weeks ago

screen kept garbling my sessions and tmux felt like too much. So I built boo, a screen-style multiplexer on @mitchellh's libghostty. Sessions survive disconnects, a TUI to hop between them, and agents keep running when you close the terminal. github.com/coder/boo

8 4 44 9K 39

View Details

Steve Ruiz @steveruizok

2 weeks ago

@threepointone Yeah, I miss busy idleness, like fixing types or small refactors, that let me think about other things while still being engaged with the material

7 3 112 4K 4

View Details

Jürgen Leschner @jldec

2 weeks ago

🚀

Astro @astrodotbuild

2 weeks ago

A beta release of v7 is out now. Try it with npx @astrojs/upgrade beta

2 1 61 8K 2

0 0 1 50 0

View Details

Matt Pocock @mattpocockuk

3 weeks ago

The /teach skill might be the coolest thing I've ever built. I've been learning: - How to solve a rubix cube (I can do it now!) - Better ways to introduce my toddler to new foods - How to orchestrate vocal harmonies for folk music Works super well for each one.

65 60 2K 102K 1K

View Details

Jürgen Leschner @jldec

3 weeks ago

Agent learning should go on the balance sheet podcasts.apple.com/us/podcast/no-…

0 0 0 33 0

View Details

Jack Herrington @jherr

3 weeks ago

@FredKSchott @0u0i3 This guide is so cool. As a visual learner I wish there was more stuff like this.

2 1 10 1K 10

View Details

Jürgen Leschner @jldec

3 weeks ago

lfg

Boshen @boshen_c

3 weeks ago

Ok I'm starting this new project two hours after joining Cloudflare. Wish me luck!

29 15 867 79K 41

0 0 2 110 0

View Details

Jürgen Leschner @jldec

3 weeks ago

Ooooh nice 😍

Cloudflare Changelog @CFchangelog

3 weeks ago

Cloudflare Workflows now supports Saga-style rollbacks. Add compensating logic to each step to handle cleanup automatically. developers.cloudflare.com/changelog/post…