Botting Findings

As I said, we are a group of people. We communicate and share various projects with each other. We are not robots. There are often transfers between our wallets. At first, only a few of us got together because of our common hobbies. Now we have dozens of people. Are we wrong to learn from each other and grow up.

For a period of time, the test coins could not be received normally, so we transferred some test coins to each other. At that time, I also consulted many times in the discord group. Why can’t I receive the test coins normally? Unfortunately, no one helped me solve this problem. You can go to see my discord chat records. My discord user name is Haitang.

Isn’t it OK to exclude robots? Why should we reduce our.

It would be easily possible to see this imho, if the trades didn’t occur in the same fashion (testing the same assets and features) and didn’t occur with the same time periods in-between. Is that possible to see @kilpatrick?

I agree with this part, I don’t think there is a group of people “helping” each other trade free faucet tokens, or a group of people using dapps one by one,Those who say they are helping each other can actually be seen to be bots, by simply checking the address of the main network

The problem is that, for example, the first chain of 4 addresses here haven’t done the same actions, but some variations of them - different assets and different quantities.

They did occur, as stated above, within a 1 hour period from each other though, which should be enough from what we’ve discussed? DAO should decide here.

Here’s the cluster on mainnet.

@haitang I believe it’s possible that some of this cluster is a group of friends but some of the friends are still are farming the test in a pattern. For example here’s 6 addresses chain from your mainnet cluster that were flagged for sybil farming:

The tests are proper tests, like @kilpatrick said different assets tested in different quantities, but it’s nearly impossible to get that distribution in the start and end times and break between tests unless a single person is doing the testing on these addresses, so that they test on one address and then quickly switch address and test on another address. Or maybe I’m wrong, but if so, I’m interested in hearing why your group members decided to coordinate the testing in these non overlapping chains with a few minute breaks in between.

As a side note, both haitang.eth (0x69e258c76f3aa94f28478040ea0705ebafdb1ddf) and haitangkeji.eth (0x2917634fd7319c38d377fff1d9ccac4505921339) are in the cluster.

In my opinion, it’s not even relevant to argue whether some sybil farming was done within this cluster. The relevant discussion is whether the DAO should slash sybil farming when detected, given the data.


There’s nothing wrong with learning from each other. But we don’t talk about learning here. We talk about that your group of friends demonstrates onchain behavior patterns which can be more easily explained by “a single user with multiply addresses” thesis, than “multiply users with multiply addresses” one. And the former case has pretty little value for the project.

Yup, I’ve seen your acc in Discord, you’re pretty active in CN group (I can’t read Chinese, tho). I only don’t know why you started to complain about the in-app faucet on Dec 5, when your group was already done mostly.

E.g. lookee here
On Dec 3 the addy successfully claimed the in-app faucet (dai, usdc, wbtc), then in 5 mins had got 5 incoming eth tx from 5 different addys, all within 1 minute, had like 20 mins of activity, and then passed the eth forth. Wtf was that? What your group was supposed to learn? Idk.

1 Like

Did you participate in the early test? Or dis early contribution? If you are not involved in the above, what are you doing now? If the project has no value, will you still do this? This is a good Dao. It’s not Conan in the cartoon. I’m very bored about this. Why don’t you think about how to do token Economics well and how to do marketing instead of doing something worthless here. At least for now, I think it’s a cartoon Conan

I agree with everything here.

Maybe this cluster (haitang’s) can be slashed 2 times, instead of 3 mostly due to haitang’s resilience and the obvious language barrier and timezone difference.

If accepted, it will be a 1-off exception, because haitang and his group have been vocal for a long time. Any new similar requests shouldn’t be considered and all other established slashing criteria should stay the same, in my opinion.

1 Like

Let’s finalize this matter as it becomes a mental burden for @kilpatrick and @spadefish who have done an amazing job so far. There was plenty of time to respond and correct things, and as I see the list of slashed wallets is minimal anyway, so there is no “detrimental” policing or anything, just literal clusters. The goal is not to rob the community at all, but just remove the largest clusters of like 7/10+ wallets. Thank you guys for doing tremendous job! This isn’t part of the DAO work per se because technically speaking it’s a distribution (the entity which existed before the DAO) but we wanted to have community say in it. And imho, it turned out great! So this doesn’t need to go up for voting because it can’t really (nor is there really anything to vote on).

As such, let’s:

  • Cut the clusters by 5 (so if you had to receive 100K tokens you receive 20K). As such, clusters still get a good share of the network and they did the work, but their effect is minimized.
  • As @kilpatrick suggested, haitang’s case case be slashed by 3 only and not 5.
  • The case of @AliLeymo can also be slashed by 3 only, not 5.
  • For the rest, general 5 applies.
  • The sers get a 500,000 GEAR each for the amazing work as a pre-DAO task by ex-core (not DAO distribution), and hope they will now focus on DAO analytics work! There we a few other sers who helped with things but were a bit less active or did like 2-3 replies yet valuable notes - so for those people, it’s up to you @kilpatrick and @spadefish if you wanna share some, because you led this.

Let’s benefit the protocol now, enough policing :slight_smile:


I’d actually still prefer 10% percent of the 133 address cluster, which we all agreed should be slashed in full as it’s absolutely obvious botting. Would equal a bit less than 400K for me.

For @spadefish, he did enormous work in the past few weeks and the equivalent of 1 Discord Max reward isn’t really fair either to all the things he did when compared to the Discord reward criteria of a few constructive messages.

@ivangbi If we can only get part of the Discord rewards (or funds provided by ex-core) as early task compensation for whatever reason, I’d advise for it to be double, e.g. 500k each for spadefish and I if everyone agrees.


Clear, fair, edited. Imho don’t ask for “everyone agrees” as we will get discussions on a topic that was pre-DAO (so technically speaking it’s pre-DAO business, the last piece of it). Good now?


I’ll share a 100k of my reward between the rest of the less active searchers that have done their own unique analysis and whose found addresses are included in the “Addresses to be slashed sheet”.


Should we get some compensation for our reply and analysis in the post? At least we’re involved? Is this more fair?

Ok lovely! Then please post the final list and let’s call it donezo

Great news @ivangbi! I’m very happy with the proposed reward and will also spread 20k each for @Matteww, @rotorooter , @Dr.Morty , @denden1010102 and @alnash who’s findings or comments I found very helpful.

The final list is here on the sheet “Addresses to be slashed”.

I added a column slash_amount that indicates the amount of slashing. As agreed, I set the obvious botter addresses (from @kilpatrick list) to “Remove from distribution completely”. @haitang cluster I set to 3x and all other addresses to 5x.

For the record, case @AliLeymo dodges the bullet here because I didn’t have time to fully parse and analyze the data. When someone uses or disperse the addresses which were used to test Gearbox are not directly connected with each other, but through the multisender contract (which didn’t test Gearbox, so the connection gets lost in my analysis).


I can’t say I fully agree with the “slashing” coefficients :rofl: But still I’m very happy that the case can be settled somehow at last, it was really a kinda tiresome and unpleasant case.
Thank you @kilpatrick , @spadefish , and other researchers of course, it’s an amazing and valuable job that you’ve done.


Well done, Thanks for your effort @kilpatrick , @spadefish. next let’s focus on token model.


great job.

Could you make a separate Gsheet with final Airdrop list

aka address <> number of tokens <> reason of slashing (if applicable)

Sure thing, I will do that during the weekend when I get the chance.