Mobilis in mobili

@noiseto_signal Gotta pick the right tool for the job Too much shiny object syndrome is what stops people from actually shipping

The most viral AI agent post this month had 27k likes. The agent ranked last in every real user review. That's the March 2026 problem. Everyone switching to agents. Most are switching to whatever just went viral. I spent the last two weeks going deep on three agents and running them through real production use. Here's what I found. Manus handles your machine. File ops, browser control and multi-step tasks. No setup pain. Starts returning time immediately. OpenClaw runs 24/7. Connects to 850+ tools. Highest ceiling but you're training it like a junior hire, not flipping a switch. Claude Dispatch controls from your phone. One persistent thread. Walk away. Feels the most like a real coworker. Save this before you pick the wrong one.

@YashasEdu People treat prompting like vibes and then blame the model for generic output… Garbage brief in = garbage post out

Most people blame AI for bad writing. The model isn't the problem. Your instructions are. 90% of content marketers use AI writing tools daily. The AI writing market hit $1.77B in 2025 but almost nobody builds the one thing that determines whether the output sounds like them or sounds like everyone else. A style guide. Not write in a friendly tone. That does nothing. A real operating manual like the specific patterns that make YOUR writing yours. Here's what actually goes into one 👇 1. Voice boundaries ‣ Direct but not arrogant, funny but not performative are constraints which shape these output ‣ Vague adjectives like smart and conversational don't change model behavior 2. Structure templates ‣ How you open, how you move, how you close makes it explicit ‣ AI infers structure poorly. If you don't specify it, you get walls of text 3. Sentence level patterns ‣ Your rhythm, your punctuation habits, how much abstraction you tolerate ‣ The more specific you are here, the less editing you do later 4. A blacklist ‣ Most useful section and most people skip it ‣ Every model has defaults like not X, but Y constructions, hedge words, summary endings. Name them and remove them 5. Positive AND negative examples ‣ Show the model a paragraph that nails your voice. Then show one that sounds like AI ‣ The gap between those two is your style guide in practice 6. A revision checklist ‣ Does this sound like a person or a polished summary machine? ‣ If you can't answer that, the draft isn't done 86% of marketers already edit AI drafts before publishing. Platform features quietly caught up too... ➥ @claudeai has projects + styles + skills ➥ @ChatGPTapp has custom GPTs with persistent memory ➥ @GeminiApp has gems where ou build style guides into workspaces now Build the system or keep fixing the same problems on every draft.

@M_zekks The “community-led growth” to ads manager pipeline 😬

Projects that had all the traction during Infofi are now making use of Twitter ads Very sad to see

@YashasEdu The narrative is decentralized, but the cash flow is concentrated

28 DeFi protocols now have active token rights like buybacks, fee switches, revenue distribution to holders. It might sound like a broad shift but it's not. As 2 protocols generate 80% of all holder revenue in the entire token rights universe. Hyperliquid and PumpFun combined push $1.15B annualized. The other 26 has about $290M together. Here's what the top of the table looks like 👇 1. Hyperliquid' $HYPE | $687M/yr to holders ‣ 97% of trading fees flow into HYPE buybacks through the Assistance Fund ‣ Over $54M in fees just from last 30D ‣ RWA perps through HIP-3 are driving volume that didn't exist 6 months ago 2. PumpFun' $PUMP | $468M/yr at a 1.14x P/F ratio ‣ Cheapest large-revenue protocol by that metric ‣ But revenue dropped 12% in 30 days, a $500M lawsuit is pending, and team wallets moved $10M to exchanges in February. 3. Aerodrome' $AERO | $78M/yr ‣ 100% of fees go to veAERO lockers ‣ One of only 4 protocols that pay direct dividends instead of indirect buyback support 4. Aave' $AAVE | $77M/yr ‣ Running a $1M/week buyback program since early 2025 ‣ Only ~13% of total protocol fees actually reach holders though ‣ The rest funds operations 5. Jupiter' $JUP | $59M/yr ‣ 50% of fees go to JUP buybacks, locked for 3 years Now 25 out of 28 protocols chose buybacks. Only 4 pay direct dividends: @AerodromeFi @pendle_fi @CurveFinance @ethena Remember buybacks let the team control timing and execution. You're trusting that reduced supply eventually shows up in price. Dividends put revenue in your wallet. One is indirect. The other is cash flow you can see but the real gap isn't even the mechanism. It's the pass-through rate. Most tokens market themselves as "fee switch ON." Here's what actually flows through → Aerodrome: 100% → Hyperliquid: ~89% → PumpFun: ~47% → Aave: ~13% → Uniswap: ~2% The switch is on. For most, the flow is a trickle. If you evaluate tokens based on does it have a buyback program you're asking the wrong question. The right one is what percentage of protocol revenue actually reaches holders and is that revenue growing or shrinking? Right now nearly every protocol on this list shows -10% to -50% fee decline over 30 days. Not protocol-specific. Market-wide cooldown from January's peak. Cheapest protocols by P/F ratio rn: → @MeteoraAG: 0.42x → @eulerfinance: 0.70x (post-hack rebuild) → @Pumpfun: 1.14x → @JupiterExchange: 1.15x → @Raydium: 1.38x ➢ Low P/F + declining revenue is a trap if the decline accelerates ➢ Low P/F + stable or recovering revenue is where the opportunity sits That distinction gets lost in the noise. One filter I've been using is to ignore the mechanism label. Look at three numbers: 1. Annualized holder revenue 2. P/F ratio 3. 30-day fee growth direction ➢ If all three line up, the token has a real value accrual story ➢ If any one breaks, the buyback is marketing Most of DeFi's token rights story is marketing. A handful of protocols are actually moving capital that matters. Know which is which before you buy the label. h/t to @DefiLlama for the data

@CryptoJonesRC @Route2FI gold cage is still a cage 🤷‍♂️ broke + free can scale… rich + trapped just buys nicer walls.

I saw this screenshot in @Route2FI’s tg channel (will link below). You should never voluntarily put yourself in a cage. It doesn’t matter what your penthouse or apartment looks like. A rich prisoner is still a prisoner. A cage is a cage. You have to go outside and experience life. Locking yourself up like this will only lead to many mental and physical problems.

n8n users writing 42-node diagrams vs Claude dispatch users sending “handle this” from the beach We are witnessing (interface) evolution in real time

GPU poverty arc is over.

Frontend devs realizing Stitch can ship 20 ui drafts before they finish renaming final_final_v3

Me to Claude at 2:43am: just 1 more prompt... Also me to Claude at 4:12am: deep in devtools logs debugging what that 1 prompt did

Poke bowl or poké ball? These guys quietly built a B2B product off the back of their B2C hit 2 monetisation paths off the same underlying data is a giga-brain move More AI models trained on niche consumer data to come… we’re just getting started

@noiseto_signal One click solution 👀

Top: Tiktok reference Bottom: Kling 3.0 motion Claude + Nano Banana + Kling 3.0 We're building the one-click version of this pipeline. Soon.

This is a feature, not a bug The era of personalised software is upon us Build what YOU need

@milalolli @gmCR3W Interested 👀

Hey London! 🇬🇧 Building in a silo is hard. The @gmCR3W community is incredible, and we want to make it even more powerful. We’re launching a member directory to help you find collaborators based on their specific skills. I'm opening up early profile creation soon. Drop a comment if you want to get in early. 🚀 It's free!

Cursor, Lovable, ChatGPT are Londonmaxxing AI summer loading

@zachtdavidson DMd

@Chewie_xo @OnChainBrits DMd

We already have 125 members Londonmaxing Dm me for an invite to @OnChainBrits - for founders, builders and hobbyists in frontiertech (AI and blockchain tech)

@Diarmuid_100x @solana DMd

gm network looking for crypto-native experts in the following: - product market fit - GTM - twitter/x marketing all in the crypto-specific context (bonus if they understand @solana’s landscape) if you know of any please tag them below opportunity to help out some hungry & talented founders 🤝

@hrkrshnn DMd

I'm seeing this a lot: high agency, fast learners with AI tools are thriving in this new era. If you're one of those people, reach out in DMs.

@milalolli 🇬🇧

If you are London based and Londonmaxxing building stuff. Drop a 🇬🇧 below 👇 I’m forming a group for IRL events and I want to see you there