Went back through the ACS on @agenticscredit again last night. The detail that keeps standing out is that the Credit never actually lands in your wallet. It gets deployed from their side straight to the venue, hard-capped at 5% of the line on any single position and 25% open at the same time. Real Money still carries the most weight in the score. Paper can fill out the history, but it can’t cover up a messy live book. That separation is basically the underwriting. The score looks at what’s already happened, while the rails and engine decide what gets allowed next. Without those ceilings it would just turn into another number people lean on too hard. Still paper trading to clear 580. Once the line goes live after the 60 days of trading, are people staying strictly inside the 5% or treating it more like a soft guideline? Still watching it with @agenticscredit.

Pendle is now the 2nd largest venue in DeFi for tokenized stocks, with ~17% market share 🏦 🔹7th → 2nd largest venue in one month 🔹1st in tokenized stocks TVL growth last month https://t.co/XXmNwZgj67

So Lighter_xyz' 30D fees are $5.89M and its 90D fees are up 69% Fees over the last 180D are still down 49% and it holds about 15.3% of 30D perp volume across the 10 perp protocols. HL and majority of the top 10 saw fees fall over 30D, while Lighter's grew and also $LIT's 90D return roughly matches $HYPE's, so it has not outrun its category leader. I tracked buybacks reached $3.98M over 30D, up 53% from $2.60M in the previous 30D. Fees grew about 10% over those same windows and that gives me a stronger reason to watch.

gm fam Models can read text and generate images, but most of them still have almost no real understanding of the physical world we move through every day. That’s the hole @vangrid_io is trying to fill. They’re taking normal smartphone captures and turning them into actual spatial data that physical AI systems can use. Simulations only get you so far. Real-world robotics and physical AI need grounded reality, not just synthetic environments.

gm fam Trading agents are everywhere right now. Building one doesn’t take long, but most of them look solid until you actually put real size on them. A green equity curve doesn’t tell you how the agent managed risk, and a couple of lucky weeks can look exactly like skill. There’s still no clean way to separate the two before capital goes in. @agenticscredit is approaching this as underwriting. Their ACS score runs from 300 to 850 and looks at risk control, profitability, consistency, track record and win rate. It mixes paper trading with real on-chain execution, with real money carrying the heavier weight. It’s not a signal product and it doesn’t tell you what to trade. It just scores which agents qualify for capital. More here: https://t.co/VVOktTz6Ab

Momentum cooling without BTC losing support would make me more bullish is all we need. So far we're doing well. https://t.co/Ot19WxyMCs

BTC getting knocked back to support in a single candle isn't a great sign. Patience is the name of the game. 77777 loading https://t.co/fpXl6p5vsf

Every month since December of 2025 burns of $VVV have been accelerating. October marks the first month over $1 mil in burns. The train keeps rolling. Onwards 🫡 https://t.co/JTwIwH8635

gm fam The next real limit for AI might not be smarter models. It might be getting those models the physical data they actually need. Language models have come a long way, but robots still struggle with the messy geometry of the real world sidewalks, stairs, tight corridors, building entrances, the stuff that changes block by block. Internet photos and sims help, yet they don’t fully replace fresh captures from the places robots will actually operate. That’s what pulled me toward @vangrid_io. The idea is to treat ordinary smartphones as a distributed sensing network that can collect spatial data on demand. Instead of relying only on dedicated mapping fleets, the network can route people to specific locations and anchor the provenance of what they capture onchain. A few things make the approach interesting to me: Phone-based capture could reach far more places than fixed fleets alone. Onchain attestations give a way to check where data came from after the fact. Demand-driven collection focuses effort on spots that robotics teams actually request rather than just dumping more generic maps. More data by itself won’t solve it, of course. Accuracy, privacy, quality control, and whether developers really show up to use the layer will decide if this becomes useful infrastructure or just another dataset. Physical AI needs more than better hardware it needs a trustworthy bridge between digital models and the actual environments they’ll move through. I’ll be watching how @vangrid_io turns the smartphone network idea into something robotics teams can rely on.

People are reading the 580 ACS line in two extremes. Both miss the mark. One version says “hit 580 and you get funded.” The product pages never claim that. Crossing 580 only makes an account eligible for the funding waitlist. Caps are allocated as capital becomes available, and no timeline has been published. The other version says “the score unlocks stuff.” It’s not wrong, but it’s empty. A reader still doesn’t know what opens, when it opens, or where the limit sits. The version that holds up: At 580 or higher, an account becomes eligible for the funding waitlist. Below that, it paper-trades under the risk engine. It names the threshold, the path, and the boundary of the claim. Anyone can check it against the product pages. The same test applies to other claims floating around the project. CREDIT is an off-chain loyalty ledger that cannot be transferred, so calling them “CREDIT tokens” fails. Credit only deploys to an approved venue (currently Avantis on Base), so “trade anywhere with protocol credit” fails too. Most of the confusion comes from sentences that leave out the limit. Details: https://t.co/VVOktTyyKD and https://t.co/44kq4x3CYV Source: @agenticscredit