Tim Roughgarden pfp

Tim Roughgarden

@tr

125 Following
1210 Followers
Tim Roughgarden pfp
Tim Roughgarden
@tr
Over the past three summers @a16zcrypto we've had 19 tremendous research interns. Want to be part of the summer '25 cohort? The application is open now, for full consideration please apply by Nov 8. https://a16z.com/about/jobs/?gh_jid=6242445003
1 reply
15 recasts
54 reactions
Tim Roughgarden pfp
Tim Roughgarden
@tr
gotta stop getting my hopes up for a sunn o))) concert every time someone advertises a drone show
0 reply
79 recasts
81 reactions
divya pfp
divya
@divyav.eth
a great short intro course by @tr to learn how incentives work in CS. I really love the format - have been watching the videos whenever I have some time in between things. https://youtube.com/playlist?list=PLEGCF-WLh2RJdrKZ431SidRX_T4VmAKx8&si=23HW3dy0b2hpfBCh
0 reply
1 recast
1 reaction
a16zcrypto pfp
a16zcrypto
@a16zcrypto
LVR (“lever”) introduced a much-needed way for DeFi liquidity providers to think about whether and when to provide liquidity to an AMM. Check out the clearest explainer (in just 3 minutes) of LVR yet, from LVR co-inventor and Head of Research at a16z crypto @tr. https://www.youtube.com/watch?v=O7WoMeC0zow
1 reply
2 recasts
13 reactions
Tim Roughgarden pfp
Tim Roughgarden
@tr
Finally got around to uploading the corresponding video: https://www.youtube.com/watch?v=l8uRkvrnc_c
1 reply
6 recasts
20 reactions
Tim Roughgarden pfp
Tim Roughgarden
@tr
Thanks for the kind words! Re: future, see https://warpcast.com/tr/0x54ed0356
0 reply
0 recast
1 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
Major update to the paper with @mbahrani @pgarimidi on transaction fee mechanism design in a post-MEV world (i.e., with active block producers), now with searchers (Sections 2.5+4) and tight welfare guarantees (Section 5). Full paper at https://timroughgarden.org/papers/eobp3.pdf
0 reply
1 recast
9 reactions
Tim Roughgarden pfp
Tim Roughgarden
@tr
Independently, Yotam Gafni and Aviv Yaish proved a similar impossibility result for the case of deterministic mechanisms. Their paper is well worth reading, as their proof techniques are interesting and different from ours: https://arxiv.org/pdf/2402.08564.pdf 7/7
0 reply
0 recast
0 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
Caveat: this work considers the original, pre-MEV model for transaction fee mechanism design. Adding in MEV does change things: https://arxiv.org/pdf/2307.01686.pdf 6/7
1 reply
0 recast
0 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
Answer: no, even for randomized mechanisms! In this sense, the EIP-1559 and tipless mechanisms are "optimally incentive-compatible." 5/7
1 reply
0 recast
0 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
A variant that I called the "tipless mechanism" has the same set of properties, except that it loses OCA-proofness rather than DSIC when the base fee is too low. Obvious question: can we get all three properties, all the time, with no extra conditions? 4/7
1 reply
0 recast
0 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
EIP-1559 satisfies all three properties as long as the base fee is high enough that all eligible transactions fit into a single double-size (30M gas) block. But if the base fee is too low, it effectively reverts to a first-price auction and loses DSIC 3/7
1 reply
0 recast
0 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
DSIC: users have an "obvious optimal bid" MMIC: miner/validator is instructed to maximize their fee revenue, has no incentive to insert fake transactions OCA-proof: if miner + all users collude off-chain, they can't do better collectively than in canonical on-chain outcome 2/7
1 reply
0 recast
0 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
Happy to report that the main open theory question from my original work on EIP-1559 has been resolved (with two brilliant collaborators, Hao Chung and Elaine Shi)---no transaction fee mechanism can be DSIC, MMIC, and OCA-proof! https://arxiv.org/pdf/2402.09321.pdf (more context below) 1/7
1 reply
0 recast
12 reactions
Tim Roughgarden pfp
Tim Roughgarden
@tr
not familiar with it
0 reply
0 recast
0 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
True even the most privileged don't currently have too many other avenues for learning that material in a time-efficient way...
1 reply
0 recast
1 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
One unexpected thing about the MOOCs is that it's in some sense the "pursest" teaching that I've ever done --- no certification, no assessment (other than self-assessment), etc. Everyone there was/is hungry for knowledge and skills, in many cases with few other avenues to attain them
2 replies
1 recast
6 reactions
Tim Roughgarden pfp
Tim Roughgarden
@tr
But expect a lot of educational materials from me (as always) as this decade unfolds (Bur right now, research and other responsibilities call ;)
0 reply
0 recast
1 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
That was an ephemeral moment --- no recordings, and I'm a white/blackboard lecturer so no slides
1 reply
0 recast
1 reaction
Tim Roughgarden pfp
Tim Roughgarden
@tr
I've gotten many hundreds of messages like this since my MOOCs on algorithms launched in 2011. Hearing stories like this never, ever gets old. @has so happy you found it useful!
1 reply
1 recast
10 reactions