Griffin Anderson 🟠 (@andergri) header
Griffin Anderson 🟠 (@andergri) profile picture

Founder @phi_labs, core contributor to @archwayHQ Social Disclaimer: https://nitter.cf/t.co/kCXPDJLnCs

Joined August 2010
Photo by Griffin Anderson 🟠 (@andergri)Photo by Griffin Anderson 🟠 (@andergri)Photo by Griffin Anderson 🟠 (@andergri)Photo by Griffin Anderson 🟠 (@andergri)Photo by Griffin Anderson 🟠 (@andergri)Photo by Griffin Anderson 🟠 (@andergri)Photo by Griffin Anderson 🟠 (@andergri)Photo by Griffin Anderson 🟠 (@andergri)Photo by Griffin Anderson 🟠 (@andergri)Photo by Griffin Anderson 🟠 (@andergri)
Griffin Anderson 🟠 retweeted
What if 80% of DeFi liquidity is sitting idle? Max McKendry (@CryptoChem0000) from @BoltLiquidity explains how prop AMMs can support trading volume with less capital tied up.
2
1
5
368
Griffin Anderson 🟠 retweeted
There are only a handful of ways to make RWAs actually trade onchain Our CEO @andergri on why L1s and L2s are suddenly hunting for propAMMs πŸ‘‡
4
7
1
14
566
Griffin Anderson 🟠 retweeted
I don't think that people are realizing how genuinely efficient propAMMs are compared to traditional ones $21M of TVL for BisonFi is doing more monthly volume than Raydium and Orca on Solana is wild
1
4
182
Griffin Anderson 🟠 retweeted
PropAMMs went from 0% to ~half of Solana DEX volume in a year. β€’ 0.26 bps spreads vs 2.59 on classic AMMs β€’ $21M TVL doing $8B+/month β€’ 98% of volume in one pair β€’ Net economics: unpublished Why it wins, where it breaks, and what's next πŸ‘‡
8
12
2
19
1,663
Griffin Anderson 🟠 retweeted
Nearly $40B in RWAs onchain and still climbing. Next step is deeper markets to match. We're built for that.
there’s no stopping tokenization
6
7
14
364
Chainalysis: in late 2025 the median x402 wallet was 73 days old at its first payment. By Q3 2026 it was 18 minutes, and 67% of new wallets paid within an hour of funding, up from 4% a year earlier. Agent wallets have become expense cards. The top-up is the control that matters.
132
Bittensor's eTAO lets owners rewrite the fixed 18/41/41 emission split. Owners can take up to 33% and cut miners to 33%. Root TAO stakers keep their default share automatically. Alpha validators absorb the change. Root got a guarantee. Validators got the variance.
1
1
148
Coinbase now lets Claude, ChatGPT and Grok trade its index and crypto perps. The fine print: "Agents are not offered by CFM," its CFTC-registered futures arm, and you "assume all risk" for what your agents execute. The leverage is regulated. The trader belongs to nobody.
1
2
145
NEAR AI's Lean agent solved 672 PutnamBench problems, averaging about $0.17 each. Next target: formally verifying the NEAR protocol and its core smart contracts. Machine proofs are getting cheap. The spec isn't. Proving the wrong rule is now cheap too.
1
1
148
Griffin Anderson 🟠 retweeted
Token2049 felt like an inflection point. The market is catching up to the problem Bolt was built to solve. CEO Griffin on why the timing feels different now πŸ‘‡
5
5
2
18
898
Prime Intellect built prime-rl for INTELLECT-2, a permissionless RL run that four RTX 3090s could join. Today's README pitches 1,000+ GPUs on Slurm and Kubernetes and never says decentralized. Async RL was sold as the decentralized unlock. It shipped as a datacenter feature.
1
1
147
PRXVT's first px402 hid x402 agent payments with zero-knowledge proofs: 2+ seconds per payment, over half an hour of proving for 1,000 calls. v2 switched to blind signatures, about 0.0004 seconds of wallet work each. Agent privacy goes to the fastest scheme, not the purest.
1
1
1
123
Polygon on Kira, an early Open Money Stack customer: AI agents handle account opening, KYB, RFIs and payment tracking across U.S. banking and stablecoins. Humans sign off on every decision. The agent gets the compliance work. The signature, and the liability, stays human.
3
2
1
16
1,433
February: NEAR AI said Agent Market showed contracts and courts getting replaced with AI agents coordinating. October: payouts via Stripe Connect or USDC, an automated resolver about every 12 hours, human review after any ruling. Agent commerce shipped the court back in.
1
2
168
Brave Leo now runs GLM 5.3 Flash in NEAR AI's TEEs, end-to-end encrypted, in every release channel. Brave's test: 297 microseconds to encrypt the prompt, 5.76 ms to decrypt the reply, on a 19.9-second request. The encryption is a rounding error. Unattested AI is now a choice.
2
3
215
Bittensor's new Null Consensus lifts a subnet from 256 to 2,500 participants. The price: the participant with the highest stake weight becomes the only validator allowed to submit weights. Opt-in; Yuma stays default. Scale was bought by naming one judge. Watch who opts in.
1
1
225
Dimon on JPMorgan's Q3 call: the bank won't "allow just any outside agent" in, and customers who let agents in should understand the "liability transfer." Banks aren't refusing agents. They're refusing to be liable for them. Whoever underwrites the agent gets the customer.
1
163
Headlines said Hedera launched Agent Accounts for AI agents. Hedera's own post: limited-partner testnet in Q4, public testnet in Q1 2027, mainnet targeted for Q2 2027 pending security review and legal sign-off. In agent payments, "launched" now means the early-access form opened.
2
1
228
Grayscale's AI crypto sector returned 54% in September against 24% for crypto overall. NEAR did 183%, Venice 70%, World 47%, Bittensor 37%. Until agent payment volume moves the same way, a 54% month is a positioning trade, not a usage story.
2
2
206
Oobit's new AI Payments: you approve a purchase and a spending limit, and the assistant gets a one-time virtual card funded from your stablecoin balance. It never touches the USDT. In agent commerce, stablecoins are turning into the funding account, not the rail.
2
2
132