Just released evmstate, a TypeScript library to trace & visualize state changes on EVM chains, with human-readable labels (variable names, mapping keys, decoded values, etc).

RT @FUCORY: If you are a cracked rust, c, c++, or zig dev who likes working autonomously and shipping top tier code reach out.

RT @adamscochran: Anyone who has done an ounce of work with AI, even as a user, can tell you that this system prompt change will *NOT* sudd….

@grok what do you think about this claim?.

The commit: thanks for sharing @bantg.

The system prompt they added to grok includes:.- "[. ] Assume subjective viewpoints sourced from the media are biased.".- "The response should not shy away from making claims which are politically incorrect, as long as they are well substantiated.". Important words being "the.

Guy that is obsessed about truth-seeking and free speech cannot stand an AI trained on facts speaking ill about him and bans people from the social media he bought when they expose him.

Norway and Denmark, two social-democratic countries, have among the highest Human Development Index scores in human history.

Super promising. I listen to techno/trance and fragmentation mostly between spotify/soundcloud/youtube is a real pain.

Figured out why this pairs this well with iterating with a mentor or anyone with more experience, where they will notice the smells and pinpoint the right design/api, and help you iterate. Aka how you build up your intuition. The LLM is the tool for me to notice outstanding.

I was doing that this week , started blindly with a rough idea of the desired output, and the repo became huge with a bunch of dead code and trying to integrate too much breaking stuff at once. Now I know exactly what I want it to look, so I deleted everything and wrote a prompt.

Something I like to do with a coding assistant, and that works great is:.- vibe code your app or whatever you want to do; prompt as best as you can, give some references, documentation.- you don't have the correct design/structure in mind so LLM will mess it up, produce a bunch.

I’d say claude code max is the best value out there + you can login to the gemini cli with a google account for the free daily requests, so you can tell claude to use it in non interactive mode to summarize large contexts. You can just pay the $100 and upgrade to $200 if you hit.

RT @HappyChainDevs: Yeet! The HappyChain testnet & the Happy Wallet are here!. ‣ Log in with Google.‣ Install nothing.‣ *Real* cross-app wa….

Entrepreneur mindset is so cooked. I get the take on "just build stuff", but turning that into your entire take on college is wild. I know this post is just the usual linkedin roleplay and there is no actual thought process behind, but still this is something I see a lot. Any.

Maybe that's old news but I hadn't noticed. If I understand correctly, anyone can sign a new authorization and send an EIP-7702 tx to change the delegated contract.

The "AdvancedCrimeEnjoyer" contract is basically:.- `receive()` → auto-forward ETH.- `fallback()` → auto-forward ETH.- `transfer()` → auto-forward tokens.- `execute()` → only the delegator can call. Contract:

`getCode` on the test address now returns `0xef01<delegated_address>`. Delegated address being this "AdvancedCrimeEnjoyor" contract: Meaning that a call to that anvil account will execute the code of that "AdvancedCrimeEnjoyor" contract.

Just noticed that with EIP-7702 someone delegated anvil's default test addresses execution to a malicious contract to automate the bot sweeping. Now any transaction triggers contract code to redirect the funds. Final destination:

RT @0xDesigner: 13 prompting techniques for better vibe coding. small changes that yield massive performance improvements 🧵👇.

RT @adamscochran: This is why the danger isn’t true AI/AGI. It’s humans. People like Musk with perverse agenda’s will attempt to make AIs….