eng @coinbase working on @coinbasewallet | building products that give people more agency | @withcortex
NYC
Joined November 2018
- Tweets5.9K
- Following4.4K
- Followers2.3K
- Likes81.3K
GM from @cursor_ai Compile in NYC! My first takeaway is there’s a good deal more I could be doing to parallelize my work with Grok Bot
If you watch one thing today I’d recommend you watch this talk, really interesting points & potential glimpse into the future of software development
How Amp Builds Amp
2026-10-13 • @paradigm Frontiers
youtube.com/watch?v=kQguG4YS…
5
344
Really good watch from the All In Summit
youtu.be/A4Q7zAayW20?is=5Ctz…
10
1,184
@OpenRouter's free model router is very under-utilized, highly recommend to any devs looking to add a simple/free model option to their app(espc. where frontier intelligence isn't required)
openrouter.ai/openrouter/fre…
1
451
Seeing all these old Apple presentations on my YT feed, just watched this one unveiling the first Apple Watch
Always so inspiring to hear about all the craft that goes into net new interfaces/devices and how prime Apple is able to wow the audience with every detail of their presentations so well. Also miss seeing the live presentations & crowd reactions at these Apple events! All very inspiring to see, can’t stop thinking about it
youtu.be/5uo5bpDYtIE?is=viZ4…
1
3
872
What I like about running the agent on a control plane is you can get better durability and can more elegantly handle the state of the VM — the VM is just an execution environment and even if you need to restart it you can feed that agent logs from the last session etc
But what I like about running the agent on a VM is that you can just SSH into any machine to set this up *and* it seems better suited to easily run any harness out of the box. Otherwise it might be a bit harder to configure the harnesses tools to use the remote machine.
Not too sure which one of these is the best approach, and I’d want to lean on the side of interoperability(easily switching out harnesses or switching from a cloud machine to a local machine), but would be curious to hear from other folks tinkering with this for their own needs 🙏
2
252
Welcome back @coinbasewallet 🫡
We’ve been working hard this year to make the best onchain trading experience. Download the latest update to see what the team’s been cooking up!
Really cool stuff going on rn with Minimax H3 Max, scratching the surface of the types of entertainment(and interfaces) a video model like this can afford
Minimax H3 Max has generates video faster than you can watch it so I hooked it to a twitch livestream! Now you can watch infinite interdimensional cable - link to the stream below
2
7
977
Riding the wave and installed Omarchy on an old Mac Mini, really excited about it so far
I’ve tried repurposing this machine many times but it all felt too slow or complicated, the Omarchy setup was super fast and I’m enjoying playing around with a new tiling manager/way to use the desktop. Fun times!
New blog post: When you give an agent a computah
Goes over how we went from early tool calling models to coding agent harnesses with their own computers, what the right execution environment for agents might be(eg. remote machines vs v8 isolates), and the new interfaces that come after the IDE & desktop metaphor
xitter.cf/Dylan_Steck/status/208…
Also available to read on my site
dylansteck.com/blog/when-you…
4
137
Can't say enough about how cool I think it is both that zSwap fully runs on the blockchain and that Freedom Browser can pull content directly from the blockchain without intermediaries - this is the kind of stuff that made me want to get into crypto in the first place
👏 @heckerhut & freedom browser
this is as ethereum as the internet gets
freedombrowser.eth.limo/
2
11
850
Working on a new blog post for my personal site that's powered my Markdown files, and in the past I would've wanted a custom visual editor for my site but now I just edit the file with a live preview of my app running in the T3 Code desktop app
Feels good! New post soon 👀
3
16
621
I think Fable’s awesome but I’d suspect part of the reason for this is that there are a lot more cost effective models that can do most(or all) of what you need - including open source models!
Grok’s new 4.6 model is a great example, really strong and relatively cheap model
NEW from Ramp AI Index: disappointing adoption of Fable 5
We've heard several reasons from businesses...mainly Fable 5 is just too expensive.
A model so powerful it was briefly banned, and yet businesses don't think it's worth the price.
5
9
753
Damn, if AWS engineers are having a hard time getting EC2 instances then we’re all in for a rude awakening
AWS is telling engineers to cut CPU waste as demand strains its EC2 capacity. Some engineers now wait days for servers they used to get within hours.
Read more: thein.fo/4w9oRn3
3
547
One day into using native-sdk.dev by @vercel for a side project, can confirm it is indeed goated
2
12
1,392