Abdullah Alatas

NEOBANK TIKTOK TIER LIST: I ranked the TikTok game of 10 neobanks from S to D. Most of them are failing for the exact same reason. If you're running a neobank, you should bookmark this thread.

launch week numbers don't tell you if a campaign worked. month two does new coke beat pepsi in 200,000 taste tests and still got pulled after 79 days crypto campaigns get judged on week one. impressions, signups, a volume spike. but nobody really checks if those traders came back we found this out paying communities to trade. some go hard for a week because they're getting paid, then they're gone. others just keep trading month after month. you can't tell them apart until month two so before we put real budget behind a community, we ask if they'll still be trading after month one one new but easy test before you renew: pull your launch week users and count how many traded again this month

launch week numbers don't tell you if a campaign worked. month two does new coke beat pepsi in 200,000 taste tests and still got pulled after 79 days crypto campaigns get judged on week one. impressions, signups, a volume spike. but nobody really checks if those traders came back we found this out paying communities to trade. some go hard for a week because they're getting paid, then they're gone. others just keep trading month after month. you can't tell them apart until month two so before we put real budget behind a community, we ask if they'll still be trading after month one one new but easy test before you renew: pull your launch week users and count how many traded again this month

outreach got so automated that I'm pretty sure half the bots are just DMing each other now if you DM me and it's actually you typing, I'll probably reply out of pure shock

Greg's right about SaaS. The same thing is happening to marketing. Nobody really wants to pay for the dashboard. They want the result the dashboard was supposed to show. We felt it first on the buying side. When we book communities for a campaign, the usual deal is a flat fee up front and a promise. Sometimes the room moves. Often it doesn't, and you've already paid. So we changed how we pay. A lot of our community budget now goes out on target: the community gets paid after it hits the volume we agreed on. Then we put one bigger bet on a community we trust to bring real quality, and spread the rest across the pay-on-target ones. It's not clean. The big groups push back and want a retainer, which is fair, and that's where the one bigger bet goes. Some rooms hit the number with traders who don't stick around past month one. So you still have to pick on purpose. But the conversation changed. We stopped arguing about reach and started talking about who actually traded. If an agent can do the busywork for almost nothing, what's left to charge for is the outcome. Price around the work getting done, not the tool and not the hours. If you sell growth, ask what you'd charge if you only got paid when it worked.

Btw, still hiring a BD Lead at Growgami to run cold outbound and close deals. Remote, no CV, apply by Oct 30 forms.growgami.com/careers/b…

People can tell when nobody actually wrote to them. Copy-paste outreach got ignored before agents too. Agents just made it cheaper to send more of it. What works for us is the opposite. Fewer people, picked on purpose, and a message that opens a conversation instead of pitching. Let the agent find and sort. But you should write the first line. Always

If an agent is doing the buying, your ads don't reach it. Greg calls Amazon and Delta blocking agents stage 1. It won't stay there. Once agents research and shop for people, the agent is your audience. It reads your site, your docs, your public proof. It never sees your banner. AI search is the first thing we pitch clients now. Getting an AI to find you and read you right is distribution. Easy test: ask ChatGPT what you do and why you're legit. If the answer is weak, fix that before buying more ads.

Honest question after that clip of someone forwarding their boss's instructions to Claude and sending Claude's answer straight back: If Claude writes it and the boss approves it, who did the work?

TOKEN2049 week in Singapore, predicted: 1. The same panel, on six different stages 2. "gm" from the same people at every side event 3. Everyone complaining about standing around, while standing around 4. One good conversation at 1am that makes the whole trip worth it

"Encrypted NFTs" is trending on X right now. Fungo Labs says every NFT in its collection mints sealed, encrypted onchain with Zama's FHE. The contract can work with what's inside without ever reading it, and randomness decides how each one looks. Not even the team gets to choose. NFTs have always been the most public thing in crypto. I like that someone is trying the opposite. Curious whether it holds up after mint.

It's been an interesting couple of days to follow if you're building anything consumer in crypto So here are the three things we think are worth knowing > Banks and card networks are shipping stablecoins > Porsche is closing its Web3 project > Google is paying publishers for AI answers More in-depth here: t.me/theagiclub/87

Most crypto teams go looking for a vouch at the exact moment it's worth the least. Why? Early on, a vouch is the only proof you've got. No users, no volume, nothing anyone can check. So one credible person saying "I've used this, it's good" does the job your numbers will do a year from now. But most teams save the big names and the KOL budget for after the raise or right before TGE. By then there's plenty of proof around, and one more endorsement barely moves anyone. We wrote up the full version on our Telegram, including the two things that decide whether a vouch helps you or backfires t.me/theagiclub/83

Sonnet 5.5 is out! And it reportedly beats Opus 5.5 on coding benchmarks at $2/$10 per MTok. Basically half of Opus's $4/$20 Same 1M context, 30% faster than sonnet 5, and early testers are already reporting way fewer tokens per task The pricing math just changed again if you're routing agent workloads

If you are winning every impression, are you overpaying? Your dashboard is green. Your quarter lost money. Both are true. Delivery is full. Budget is spending. Acquisition cost is up maybe fifteen percent over two quarters, and everyone blames the platform, the season, or an aggressive competitor. But nobody blames their own bids. We've watched a few teams unknowingly walk right into this. They win almost every auction they enter, and winning is exactly how they end up paying more for a customer than that customer is worth. Every time your ad could show, Meta or Google runs a quick auction for that one slot. Every advertiser chasing the same person puts in a bid, and a bid is really a guess at what that person is worth to them. On most accounts the platform places that bid for you, working back from a target cost someone set months ago. It weighs the bid against how good it thinks each ad is, but a bigger bid wins more often. The person is worth roughly the same to everyone bidding. Nobody knows the real number, so everyone bids their own guess. The company that wins isn't the one that valued the customer correctly. It's the one that overestimated the most. Economists call it the winner's curse. It came out of oil lease auctions, where the firm that won a block was reliably the one that had guessed highest about what lay underneath. The winning bid was a prediction, and it was wrong in a predictable direction. Picture three companies on one segment worth about forty pounds a customer. They bid > thirty, > forty, and > fifty-five. Fifty-five wins everything, acquires at fifty-five, and reports full delivery and healthy volume. They're fifteen pounds down on every customer and nothing in the account shows it. The fix sounds backwards. Bid what a customer is actually worth to you, and let yourself lose the auctions priced above it. A win rate well under 100% isn't the account failing. It's you refusing to pay fifty-five for something worth forty. Crocs played a different game. They reported roughly $52.4 million through TikTok Shop in a year, and their chief brand officer credited a test, fail, scale mindset instead of any single hero campaign. Run enough shots and the auction stops deciding your quarter, because you're no longer betting the year on one correct bid. The distinction for your next planning cycle: > a channel you bid for is a position, and > a channel nobody can outbid you in is an asset.

everybody's making their employees post online now and honestly, smartest thing they could do. nobody trusts a company account. you see the logo talking about itself and you scroll. but the engineer who actually built the thing, or the founder who explains it like a human? you stop for that. Edelman asked 34,000 people and "my employer" scores 78% on trust vs 64% for "business" in general. nowadays, people trust a person more.