Cena DeepBook ProtocolDEEP

UNLEASH YOUR CREATIVITY WITH GROK IMAGES Discover the power of Grok, the AI built by xAI, to transform your ideas into stunning visuals. Whether you're dreaming up futuristic landscapes, whimsical characters, or professional designs, Grok's image generation turns words into art with unmatched precision and flair. Analyze uploaded images for deep insights, edit them seamlessly to perfection, and elevate your content game. Source: Grok

🧵 ARTIST SPOTLIGHT: @MaNiCArt_ I've spent the last several months researching/collecting this incredible expressionist artist in the NFT space. Here's my deep dive on his journey, the rare CMK collection (88 pieces), and why collectors should pay attention...

If you enjoyed this thread, follow me for more deep dives on emerging artists in the NFT space! And of course, go check out @MaNiCArt_'s work on @SuperRare and the CMK collection on @opensea . If you have any questions PLEASE reach out, I am here to help!

RT @demishassabis: Gemini 2.5 Deep Think is state-of-the-art performance across many challenging benchmarks!

After one week of hacking, we had an intense day of demos with every single participants presenting what they got done and their plan for the next week We started with a presentation by @williamhwgeorge on pairwise bounded comparisons for groups, basically applying cluster QF formulas to pairs rather than groups @sarkazein7 followed it up with an application to OSS, attempting to quantify the value of contributions to git repositories @hey_stefi_ then did a deep dive into automated metrics to evaluate progress, based on which we can have a more regular stream or RPGF rewards than we currently do This livestreamed segment was rounded up by @clesaege who presented a model where traders can bet on a repos score by a judge if it was evaluated, while @juanbenet gave us a concrete use case for all these ideas: the PL dev guild that gets a new cohort every 6 weeks enabling rapid iteration of new ideas, with payouts to them depending on scores given by their mentors The 2nd half was not livestreamed, so I'm only taking names where i have permission to do so @CS_Synthesist gave a talk on a computerized system for managing people, with particular interest in condorcet voting as a formula Another participant spoke about modelling agent behavioral in these impact evaluator systems and how it changes with extended timelines and clear vs lottery reward functions The next participants then spoke about how impact evaluators become more right upon application of control theory, with a 3 step process: defining a goal or success, defining key actors (like miners in btc) along with the outcome we want (preventing double spend) and finally letting the system self organized through the fitness landscape defining the outcome but not steps required to get there Next up was a presentation on how retro rewards can ideally lead to long term cooperation instead of short term opportunism by shaping collaborative behavior Ideally we have an assess, estimate, infer loop that links impact evaluation with impact. Next was an application of this framework to create a graph of actions, run a pagerank algo to get weights & finally put into a reward function for distribution in online gaming or for impact certificates We then got a theoretical perspective from @0xNaiky on how despite being endogenous loops, impact evaluators still depend on an external factor (price) for the incentive system to work A very specific application of IEs was then proposed by @iammadab in the mathematical domain of formal verification, where a new programming language called lean allows for more seamless Collab by showing proofs without having to actually so them @bah_eth had a really fun presentation on combining pump fun memecoin dynamics with kaito style rewards for social media activity Our goal is capturing social media impulsive behavior with pump & dump schemes to create long term value for communities rather than enrichment of a few @davidgasquez then looked at how we can get better rankings from AI, highlighting issues with jury such as order of voting being important or LLMs giving everything a 7/10 He applied a pairwise model where AI can be given a list of items to compare as input and as output we get a rank between them @dwddao followed up with a provocative invitation to all participants for creating a digital twin of themselves, using just 1 hour of their speaking time in a podcast. His key insight was to decoupled opinion from judgment We then had a presentation on generalizing the @ProtocolGuild contract so that anyone can create a splits and give weights between them using any formula they prefer @owocki then spoke on solving eths biggest challenges, evolving capital allocation software & alignment between tokenholders, contributors & ecosystem, via tokenizable extractable value thru proof of allocation intelligence We then had ideas on provable data being brought onchain to measure IEs & standard structures to communicate impact (cont'd)