If you are not using Claude code this way, then you are NGMI IMO

2 replies

1 recast

3 reactions

I’ve been testing Zed with open code. The UI isn’t my favourite (I know that’s subjective), especially the agent chats which are more plain than Copilot (my favourite visual), Claude, or Codex. That said I’m going to keep testing and see if I can make it fit for me because there’s a lot to like on paper. I have a couple questions though if anyone knows: 1. How do I get prettier set up like on VS Code where it lives at the bottom and I can click it to clean up the code 2. I’m running the models that come with OpenCode and the CPU is spiking a bit, are those running locally?

0 reply

0 recast

4 reactions

I find it insane that in 2026 we still have to use hacky discriminated unions to work around a proper method dispatch and pattern matching syntax in python. Haskell and Julia have this basically solved but for some reason Typescript just refuses to implement this in a nice way. https://github.com/jfarid27/open-eth-terminal/pull/82/files#diff-6d7df05c116aabaf824d3ba99a4f266bed41b09acfedd3bd2b44925c237939c9R506-R522

1 reply

2 recasts

2 reactions

I was about to deploy my 7702 contracts, but then I did some math on my resupply.fi position and it was bleeding money badly. So that's disappointing. https://github.com/SatoshiAndKin/flashprofits-7702

0 reply

1 recast

0 reaction

ultrathinking btw

0 reply

0 recast

2 reactions

OK hear me out. Guy here who read a lot about physics, computer science, and finance. Theres a lot of programmers who don't bother to read pure computer science or physics of computing at all. The issue is a computer, your DNA, and a refrigerator are all the same type of engine: a system that requires energy to organize state, that also produces waste heat. This is the second law of thermodynamics at play. Once you see waste heat and energy, you fundamentally need to *pay* for computing. The computing costs are at worst just media storage, but the algorithms that require good experiences like compression, rankings, and other analysis require much more compute, at distributed scale. The web2 problem is fundamentally about paying for the compute costs because everyone wants these unique bespoke rich experiences, but no one wants to pay for them. Blockchains helped with this because you distribute the compute costs across various providers and pay a fee, while the mechanisms allow for verifiable compute, but this is still generally expensive. On the current UI/UX experience, the problem is algorithmic slop that's mostly advertising is generally designed to make money to pay for computing. You need algorithmic nonsense with a good mix of paid advertising to keep the lights on. There is much to be said about how bad the enshitification of social networks has become, but to run it at cost to get experiences where you see only what you care about, you're going to have to have social networks that look like email/RSS/IRC, or you're going to have to pay for it by service fees or running home infrastructure on your power bill. Nothing real-time, no gifs or videos, mostly text only. The point is the physics says it's fundamentally difficult to give people rich social network experiences without increasing compute costs, and that pushes large firms into doing more advertising and mixes of feed composition to get other people besides you to pay for it. Its not necessarily that MBAs are the problem, but that the physics forces a cost that the MBAs need to pay for, or the business won't exist.

1 reply

2 recasts

8 reactions

How does one build a client to run mini apps? Any docs on this?

1 reply

0 recast

6 reactions

Big update hitting the development branch in open-eth-terminal. I am adding agent skills to help users navigate the terminal. There is now added scripts to support claude and gemini. Other agents will be added in time since I use gemini-cli for now, but IDE and terminal agents should be able to see the new skills. Make Terminal Apps Great Again. https://github.com/jfarid27/open-eth-terminal/pull/72/files

0 reply

0 recast

3 reactions

I think I’m gonna switch from VS Code to Zed. Anyone here using it? I love Neovim and when I’m doing things in the terminal it’s what I open and I’ve become comfortable enough with vim movements and modes to know I want them to be the main input mode. In VS Code I use the Vim extension and I even use Raycast to map Caps Lock + h, j, k, l to arrow keys. But my current learning and vibe coding workflow benefits from VS Code’s feature set. From what I’m seeing Zed seems like everything I like about both of these programming environments and then some. https://youtu.be/YCPlzxshBTA

4 replies

0 recast

3 reactions

Omnipin v2 released. The CLI now uses @ensdomains Universal Resolver everywhere and `--resolver-address` argument has been removed because it is no longer needed. PR with the UR implementation: https://github.com/omnipin/omnipin/pull/106

1 reply

0 recast

7 reactions

VSCodium is a fully open source version of VS Code 👨🏻‍💻🌐 • VS Code fork that removes Microsoft proprietary tools and tracking • Supports most VS Code extensions Thanks to @nocrimez.eth for sharing this!

1 reply

1 recast

4 reactions

literally the last frontier for programmatic dynamism of the highest degree is adding vim bindings to every single text input

0 reply

1 recast

4 reactions

damn Sui is using my prof's better BGP alternative called SCION very cool

1 reply

0 recast

0 reaction

Start to be very strict with your CI/CDs, test cases and documents! There is literally zero excuse with AI in 2026. Here is how I do it: - Connect Notion MCP and GitHub MCP to Cursor / Claude - Use AI to write extensive test cases. - Use GitHub actions to Lint, Format, and run tests / build - Mandatory GitHub actions run on each Push & PR to master branch. - Use AI to document codes on Notion (with MCP) - Use AI to document each PR request before merge (with MCP) With this approach I’ve had zero failed builds, outdated docs, and unexpected failures!

0 reply

1 recast

3 reactions

Struggling to keep up with the tech landscape? Thoughtworks Tech Radar curates what's worth paying attention to. Currently, it's almost all related to AI. Even better: you can create your own custom radar to track what you want to learn and your experiences with different tech. I'm creating my own radar which will include some web3 technologies and frameworks, starting with Noir, ZK proofs, and Chainlink's DECO https://www.thoughtworks.com/radar

8 replies

12 recasts

158 reactions