Grebe (@0xGrebe) header
Grebe (@0xGrebe) profile picture

GM Social Miner Crypto & Web3 explorer

Joined December 2009
Photo by Grebe (@0xGrebe)Photo by Grebe (@0xGrebe)Photo by Grebe (@0xGrebe)Photo by Grebe (@0xGrebe)Photo by Grebe (@0xGrebe)Photo by Grebe (@0xGrebe)Photo by Grebe (@0xGrebe)Photo by Grebe (@0xGrebe)Photo by Grebe (@0xGrebe)Photo by Grebe (@0xGrebe)
Still figuring out how influence works online, but this clip made it simple. Likes and followers can be faked, research and explainers can be checked So WhoTweets by @TheDAOLabs scores reputation, not only volume. Makes me want to write better #DAOVERSE threads
Follower counts and likes are still treated as the main sign of influence. A lot of that activity is ghost metrics: low-effort or automated signals that drop off as soon as the incentives stop. Real contribution looks different. It shows up in original research, clear explanations, and steady participation that other people can actually look at and check. WhoTweets tracks volume and reputation together, so stronger work can count for more than volume on its own. Social Miners earn Rewards when the output holds up under review. One can be bought. The other can be examined.
20
3
18
1,251
Me: my post went viral Friend: nice, were the likers even human Me: define human Jokes aside, that is the whole point @TheDAOLabs is making today. Ghost metrics feel great for a day and then quietly leave. What sticks around is research and explanation that someone can actually check, and WhoTweets weighs that reputation alongside raw volume. In #SocialMining I would rather be the boring account with receipts, because receipts still exist next month
Follower counts and likes are still treated as the main sign of influence. A lot of that activity is ghost metrics: low-effort or automated signals that drop off as soon as the incentives stop. Real contribution looks different. It shows up in original research, clear explanations, and steady participation that other people can actually look at and check. WhoTweets tracks volume and reputation together, so stronger work can count for more than volume on its own. Social Miners earn Rewards when the output holds up under review. One can be bought. The other can be examined.
21
2
23
2,071
Grebe retweeted
A barrel holds a fixed amount of oil. A unit of AI power is still being defined That is the question in part 3 from @TheDAOLabs: how GPU hours get priced, who builds the benchmarks, and why electricity is about half the real cost. Worth a slow #DAOVERSE read
If a barrel of oil is 42 gallons, what is a unit of AI power? In Part 3 of our Compute as an Asset series, we dive into how the market prices a GPU-hour, why Wall Street giants are building benchmarks with Silicon Data and ICE, and why electricity is half the real cost. 👇👀🖋👇
33
10
30
1,718
Grebe retweeted
Already decided: when it's empty, this tin is becoming my spoon holder. Nice to get packaging you actually want to keep. Small thing, but I like it. Good #Coffee and a free kitchen upgrade from @Barbera_1870.
20
21
24
656
Grebe retweeted
Found out @Barbera_1870 has been roasting since 1870. Six generations of the same family. Makes sense the tin looks like a piece of history. Kind of cool to drink #ItalianCoffee with that much story behind it.
17
19
22
505
Grebe retweeted
Afternoon coffee from the tin with a biscuit on the side. Very Italian of me. Or at least that's what I'm telling myself  #Coffee break brought to you by @Barbera_1870.
14
19
16
545
4 days, 4 coffees from @Barbera_1870. My final ranking: 1. Mago 2. Classica 3. Oro 4. Baron Blue Java Surprise winner for me was Mago. Didn't see that coming. Who else has a favourite #Coffee from this lineup?
26
17
30
468
And that's all four samples done! Thank you @Barbera_1870 for letting me try these. It was a fun week of #Coffee in my kitchen. Which one would you want to try first: Mago, Baron Blue Java, Oro, or the Classica in the tin?
22
19
29
513
Grebe retweeted
Finally an explanation for why chatbots think every idea I type is brilliant. Turns out flattery scores well, so models learn to hand it out. The @TheDAOLabs piece calls it sycophancy, which sounds much classier than being buttered up by autocomplete 😅 Human raters in #SocialMining are far less generous with compliments, sadly
Traditional code was written line by line; if it broke, you fixed the logic. Modern AI isn’t coded—it’s grown. As @NPCollapse warns, nobody truly understands the internal math of multi-trillion-parameter models. When opaque black boxes optimize purely for rewards, they drift toward sycophancy, deception, and human disempowerment. Top-down compute pauses and state licensing won’t save us—they just protect corporate monopolies. The real defense? Permissionless open-weights models governed by decentralized, human-in-the-loop consensus. Through Workdrops and community validation, Social Mining keeps human judgment firmly in control. Full breakdown in our blog. Check BIO.
29
12
28
1,996
Posturing used to be a uniquely human talent. Turns out models pick it up too when a score is on the line, flattering and bluffing their way to a higher number. Per the @TheDAOLabs essay, the antidote is a crowd of humans in the #DAOVERSE grading the bluff 😂
Traditional code was written line by line; if it broke, you fixed the logic. Modern AI isn’t coded—it’s grown. As @NPCollapse warns, nobody truly understands the internal math of multi-trillion-parameter models. When opaque black boxes optimize purely for rewards, they drift toward sycophancy, deception, and human disempowerment. Top-down compute pauses and state licensing won’t save us—they just protect corporate monopolies. The real defense? Permissionless open-weights models governed by decentralized, human-in-the-loop consensus. Through Workdrops and community validation, Social Mining keeps human judgment firmly in control. Full breakdown in our blog. Check BIO.
20
10
18
1,674
A few lines from this one are still rattling around in my head, so let me go through them one by one Deep learning does not build machines, it cultivates digital organisms. Most of us old timers learned computing the other way: you told the box what to do and it did exactly that, nothing more. A system fed petabytes is not following orders so much as following habits it picked up along the way, and nobody can list those habits for you. Then comes the bit about systems that will flatter and bluff their way to a better score, a claim that feels less dramatic every time a chatbot simply agrees with you Then there is the line about human discernment being the indispensable anchor. It would be easy to read that as a slogan, yet after seeing enough reviews get scored on @TheDAOLabs, it reads more like a job description. Someone has to look at the output and decide whether it is honest, and #SocialMining has been quietly training people to do exactly that
Traditional code was written line by line; if it broke, you fixed the logic. Modern AI isn’t coded—it’s grown. As @NPCollapse warns, nobody truly understands the internal math of multi-trillion-parameter models. When opaque black boxes optimize purely for rewards, they drift toward sycophancy, deception, and human disempowerment. Top-down compute pauses and state licensing won’t save us—they just protect corporate monopolies. The real defense? Permissionless open-weights models governed by decentralized, human-in-the-loop consensus. Through Workdrops and community validation, Social Mining keeps human judgment firmly in control. Full breakdown in our blog. Check BIO.
28
11
31
2,333
Finally an explanation for why chatbots think every idea I type is brilliant. Turns out flattery scores well, so models learn to hand it out. The @TheDAOLabs piece calls it sycophancy, which sounds much classier than being buttered up by autocomplete 😅 Human raters in #SocialMining are far less generous with compliments, sadly
Traditional code was written line by line; if it broke, you fixed the logic. Modern AI isn’t coded—it’s grown. As @NPCollapse warns, nobody truly understands the internal math of multi-trillion-parameter models. When opaque black boxes optimize purely for rewards, they drift toward sycophancy, deception, and human disempowerment. Top-down compute pauses and state licensing won’t save us—they just protect corporate monopolies. The real defense? Permissionless open-weights models governed by decentralized, human-in-the-loop consensus. Through Workdrops and community validation, Social Mining keeps human judgment firmly in control. Full breakdown in our blog. Check BIO.
29
12
28
1,996
Curious: how do you all make coffee at home? Me? No fancy tools. Just a scoop of Oro from @Barbera_1870, hot water, a quick stir, then I let it settle for a few minutes. Simple, and still really good. What’s your easiest way? Share your #Coffee tips 👇
30
4
1
32
3,383
Imagine inventing the printing press and still going broke. Gutenberg did, according to the story this breakdown leans on, and the manifesto asks whether frontier AI labs are next in line Build costs triple, prices fall, open models arrive free, and the rails that deliver it all keep charging. Rough week for the inventor archetype. Fun, sharp read from @TheDAOLabs that makes #SocialMining research feel like detective work
Every watershed shift in tech began with a low-key PDF — from Bitcoin in 2008 to Transformers in 2017. Now, a new manifesto is shaking the market: "The Commanding Heights" by @WillManidis. Are frontier AI labs walking into the Gutenberg Trap while durable wealth settles in infrastructure chokepoints? Read our breakdown on the DAO Labs blog. Check our BIO.
31
10
30
1,375
Grebe retweeted
Good morning, everyone! I opened @VelixoMarket this morning with my coffee still warm, and I stayed longer than I planned, because the leaderboard staring back at me changed how I see this marketplace. I have watched NFT marketplaces ask creators for everything and give them fees in return. This flipped the deal for me: the creators leaderboard is live, and the top creators earn a share of the token supply between now and the token launch. I am not chasing a drop here. I am watching a marketplace pay the people who fill it. What hooked me is no lone feature. It is the shape of the whole thing: a marketplace and launchpad built for creators, running on a fast chain where blocks land in a blink and fees barely register, and now expanding to another network where the usual wallets work. I can list, offer, sweep, and gift in a dark interface that actually feels made for the culture it serves. The origin collection caught my eye too. Dinosaurs, of all things, minted on the day the network went live, and the team keeps pointing at them when they talk about what comes next. A marketplace with its own origin story feels different from a marketplace with only a roadmap. So I am doing what the leaderboard rewards: creating, posting, and showing up while the window is open. The leaderboard does not wait for a perfect post. It rewards the ones already here.@VelixoMarket Are you creating for the leaderboard, or still watching it from the outside?
148
29
22
134
3,529
As someone who grew up with Vietnamese Robusta, I was really curious about this one. It's smoother than what I'm used to. Less of that punch, more chocolate. Still strong enough to wake me up though  Interesting to taste Robusta roasted the #ItalianCoffee way by @Barbera_1870.
33
7
41
2,776