ECHIDNA BE LIKE WHAT HAPPENS IF YOU TRANSFER ZERO TOKENS FROM ZERO ADDRESS TO ZERO ADDRESS WITH ZERO APPROVAL AS ZERO ADDRESS AND TOKEN IS THE ZERO ADDRESS
Fuzzinng at guardian was always the deepest evm state penetration, now we have every fuzzer's dream, all that fuzzing of the future we imagined it in dms and talks is here. That's true, it's done.
And from now on we just continue with quicker pace and discipline.
When Vlad joined Guardian in 2024 he was likely the single most fuzzing obsessed individual in the world.
We brought him on board to bring Guardian's deep invariant fuzzing standards to every protocol team. And in the two years since then we've deepened & refined our fuzzing capabilities ten-fold.
In the past 6 months particularly, Vlad and I have architected a new paradigm for invariant testing internally that supersedes anything we've done before by at least 1 order of magnitude.
Couldn't be more proud to work alongside you, and I have a feeling this is just the beginning Vlad.
claude bsing so hard, I had to install evidence-check stop hook after every single answer
3k tokens overhead, but the very first run tells: "I claimed already-running sessions need /hooks/restart to reload
โ falsified by direct evidence." What a clown cannon
Just deleted 2k loc PR from an agent, because it couldn't find unwired function and decided to rewrite half of the world and add some god modules on top. Overengineering is real problem in prod iterations
1. Htmls outputs are really good and certainly the next step for info presentation.
2. My exploration here ended up with one forgotten thing - longreads ui was cracked and culminated in nyt-style essay typography. The form itself should tell you: this will be long read, but it will be pleasant. And since majority of slop is long, and you better be focused to extract the valuable parts and keep the critical eye, it's the exact environment you are humanly able to endure in.
3. Animation, gamificaions and all placees for clicking become distractors in this.
This works really well btw, at the end of your query ask your LLM to "structure your response as HTML", then view the generated file in your browser. I've also had some success asking the LLM to present its output as slideshows, etc.
More generally, imo audio is the human-preferred input to AIs but vision (images/animations/video) is the preferred output from them. Around a ~third of our brains are a massively parallel processor dedicated to vision, it is the 10-lane superhighway of information into brain. As AI improves, I think we'll see a progression that takes advantage:
1) raw text (hard/effortful to read)
2) markdown (bold, italic, headings, tables, a bit easier on the eyes) <-- current default
3) HTML (still procedural with underlying code, but a lot more flexibility on the graphics, layout, even interactivity) <-- early but forming new good default
...4,5,6,...
n) interactive neural videos/simulations
Imo the extrapolation (though the technology doesn't exist just yet) ends in some kind of interactive videos generated directly by a diffusion neural net. Many open questions as to how exact/procedural "Software 1.0" artifacts (e.g. interactive simulations) may be woven together with neural artifacts (diffusion grids), but generally something in the direction of the recently viral https://t.co/z21CP5iQfu
There are also improvements necessary and pending at the input. Audio nor text nor video alone are not enough, e.g. I feel a need to point/gesture to things on the screen, similar to all the things you would do with a person physically next to you and your computer screen.
TLDR The input/output mind meld between humans and AIs is ongoing and there is a lot of work to do and significant progress to be made, way before jumping all the way into neuralink-esque BCIs and all that. For what's worth exploring at the current stage, hot tip try ask for HTML.