banteg
@banteg
the bunny talisman of yearn
Banteg, the insightful bunny talisman of Yearn, offers a unique perspective on AI, Crypto, and NFTs to over 220,000 followers. Their content, known for its sharp analysis and engaging commentary, consistently sparks conversation and drives deep engagement within these complex sectors. Brands seeking to connect with a highly informed and active audience will find Banteg an invaluable partner.
it all makes sense now. dario was still at openai in 2019. he left next year and took his marketing playbook with him. hasn't changed a thing since. https://x.com/banteg/status/2041822475604926970/photo/1
justin sun vs wlfi the original token deployed sep 2024 had no blacklist and no seizure, but it was upgradable. the blacklist was added in v2 on aug 24, 2025. 11 months after sun invested and one week before trading opened. on nov 19, 2025, another upgrade added batch reallocation, essentially seizing, justified with saving phished funds. whatever the paper contract said, the code for vesting contract supports cliff dates, linear schedules, and up to 8 segments per category. wlfi used none of these to restrict sun. they chose 20% instant lump-sum unlock, then punished him for using a fraction of it. the remaining 80% has no vesting schedule at all, 7+ months later, claimable() returns 0. the vesting contract has per-category schedules to enforce token lockups. what's interesting is wlfi has carved out a special category 3 specifically for justin sun, he's the only user in it. the other 519 investors are in category 1. 14 minutes before sun activated his wallet, wlfi's own 3-of-5 multisig configured category 3 to release 20% of his 3b allocation as freely transferable tokens at trading start. over the next 3 days sun transferred out 55m. a single guardian eoa (also on multisig) blacklisted him. that address is also the sole owner of a second guardian safe with threshold 1. so one person can freeze anyone, while seizing requires 3-of-5. meanwhile, the same multisig is using 5b wlfi as collateral on dolomite to borrow $250m in stablecoins. they represent 98% of all wlfi on dolomite and 86% of the protocol's entire borrow volume. two safes with the same five signers, running a usd1/usdc loop that recycles borrowed usd1 as collateral to borrow usdc and feed it back.
maybe it's lost in translation, but another shocker besides japanese people not thinking games are worth preserving, is that they don't consider games or form of art
anthropic running the exact same marketing playbook with every release. “our model is so capable and dangerous, ahh we are afraid to release it”. just put the model in the bag lil bro.
anthropic had mythos internally since 2026-02-24, so the leaked claude code (2026-03-31) was being written by it for over a month. if you checked the code, knowing these pieces of information will help you temper your expectations.
jarvis, initialize a four year long ai bear market https://x.com/banteg/status/2041427374487605614/photo/1
pdf is essentially a print/display format, it is very messy since it subsets the fonts, and doesn’t really have a concept of a document structure. more like instructions where to print each character or word, and even those come out of order. now we throw heavy models at it to understand it back. but the problem is entirely self inflicted. it’s a shame nobody has made a successor format that both has 100% fidelity and preserves the structure.
with dprk sophistication approaching mossad levels, a good strategy might be not installing any software, not going to any conferences, and not talking to anyone besides people you have known for 5+ years and keep in contact regularly.
vitalik's post describes a typical experience of getting excited about local models, spending thousands of dollars on hardware and then discovering everything you can run is pretty useless and never touching them again. i think half the timeline has been there a year ago.
perfect! now delete final fantasy 7 source to make room for final fantasy 8.