Velocity Capital retweeted
#Binance is excited to announce the Mira (MIRA) HODLer Airdrop – @miranetwork $MIRA.
BNB Holders, get ready! The Airdrop page will be available on the Binance Airdrop Portal in 24 hours. Plus, this token will be listed on Binance soon!
👉 binance.com/en/support/annou…
We’re hosting a meditation event with @projectsolo and @SonicSVM tomorrow in Berkeley
August 5th, 6-8PM
Catch us there: lu.ma/j2ap386e
#SBC25
2
5
2,137
Velocity Capital retweeted
LPing on the order book has been reserved for the big players, this changes now.
Introducing Flip Orders: Concentrated Liquidity on CLOBs
Kuru brings AMM-style simplicity and control back to order books 👇
nitter.cf/KuruExchange/status/19…
Velocity Capital retweeted
Paper is out.
Link: openreview.net/pdf?id=TJjP8d…
KPOP is a new DL optimizer designed for large scale distributed training on Apple Silicon.
KPOP uses a lot more memory but is more efficient per FLOP than AdamW, so it's a better fit for hardware with a high memory:flops ratio.
Some hardware numbers:
H100: 80GB, 1000TFLOPS (fp16), memory:flop ratio = 0.08
M3 Ultra Mac Studio: 512GB, 26TFLOPS (fp16), memory:flops ratio = 19.69
The difference in memory:flops ratio is 246x!
AI algorithms are built around the limits of available hardware. Since traditional GPUs have been the main option, most algorithms prioritize FLOPS over memory.
But the unit economics of Apple Silicon favor algorithms that use more memory.
@sarahookr calls this the Hardware Lottery: often research ideas win because they suit the available software and hardware and not because the idea is superior to alternative research directions.
@tychovdo and @MattBeton are at ICML now if you're interested to learn more. We're hiring at @exolabs, all of our dm's are open!
Congrats the team @0xfravashi !!
Kuru is excited to announce our $11.6m Series A round led by @paradigm
This funding will help us scale our team and bring our vision of a fully on-chain orderbook to life on Monad.
1
1
820
Velocity Capital retweeted
"Mac Minis for example are a very good fit" - @karpathy
@karpathy shouted out my work on @exolabs in his keynote at @ycombinator AI SUS!
Here's the breakdown:
Right now most AI workloads run in the cloud where requests from different users are continuously batched together. These workloads are FLOPS-bound and favors hardware with the best unit economics of $ per FLOP, i.e. enterprise GPUs.
The personal computing revolution will shift these workloads to personal devices with lower batch sizes (mostly batch_size=1).
batch_size=1 inference is memory-bound, because all of the model parameters need to be loaded into the GPU every time a token is generated.
Apple Silicon with its Unified Memory architecture has a lot of memory and memory bandwidth per $ compared to other hardware:
- M4 Pro Mac Mini, 24GB @ 273GB/s, $58.33/GB, $5.13/GB/s
- H100, 80GB @ 3350GB/s, $625/GB, $14.93/GB/s
The unit economics of Apple Silicon are becoming more compelling with every release of the Mac. The future of AI inference looks more like open weights models (@OpenAI open weights model soonTM) run at low batch_size on personal devices.
Velocity Capital retweeted
CLOBs: Crypto’s 0-1 DeFi Moment
CLOBs are how traditional finance is finally brought on-chain. This is a $1T moment. The infrastructure is finally falling into place, and Kuru is poised to capture this once in a cycle opportunity.
A number of smart teams are also hot on the trail: @HyperliquidX, @bulletxyz, @GTE_XYZ, @hibachi_xyz, @perpltrade, @CloberDEX,@CrystalExch, to name a few.
Every team has chosen their own path while building their order book that will take DeFi from 0-1 in building the decentralized, permissionless NASDAQ alternative.
In this post I’ll shed some light on the unique design space and trade-offs to better understand the playing field, and why we believe Kuru is best positioned to win.
1. Trust and Decentralization
Many teams are building their own infra that lives outside the blockchain. This machine (sequencer/rollup) is responsible for the executing and ordering of transactions. It takes in orders and posts the batch onto the blockchain.
This means as a result you now have to trust the sequencer to handle transaction ordering properly and not censor any activity.
There are multiple levers you can pull to make things more decentralised, including zk proofs, shorter challenge windows, co-located validators, and so on.
All of these are an attempt to achieve the level of security you get with a globally-distributed L1. Not to discredit the teams working on this, but building on-chain means the state and execution live on chain.
Each transaction is secured by global consensus and is censorship resistant. This is the path we’ve chosen at Kuru.
These off-chain solutions come close, but don’t offer the same level of decentralization as our approach which we believe is key to maintaining credible neutrality.
2. Speed
Teams sacrifice decentralisation to get speed. But why optimise for speed?
CLOBs are market structures where being fast is rewarded. In tradfi being the fastest to reach the exchange has always been an advantage. This is why NASDAQ offers co-location services with fibers to maintain a level playing field.
Just like decentralisation, speed is a spectrum. In theory, a rollup's speed is limited only by what a single high-performance server can handle– the limitation is by hardware, not consensus.
In a global blockchain, things get slower and more complicated with congestion, priority fees, MEV etc. This is the price of decentralization.
L1s are pushing the boundaries with research on consensus, gossip, pipelining and much more to improve speeds for their chains. What we see today is not the final version of Monad or Solana. They will continue to improve, and it’s a bet we’re willing to take at Kuru.
Decentralization is worth the cost.
3. Composability and DeFi
Some CLOBs are focused on spot markets like Kuru, while others are building perps like Bullet.
Projects building on rollups or custom sequencers are giving up composability. Perps can live in isolation. Spot markets cannot. Composable spot markets are crucial for DeFi activity on a given chain.
Exchanges launching their own chains see the potential in DeFi beyond just order books. BSC by Binance, Base by Coinbase, HyperEVM by HyperCore, and JupNet by Jupiter. They are all making a bet on DeFi with interoperable applications.
The shared vision is still DeFi. Monad is focused on building the base layer and supporting breakout apps.
Everyone is still hyper-focused on composable DeFi.
Monad is building the infra and wants successful apps iterating on top. Hyperliquid built a gold-standard exchange and is now expanding to infra. The end goal is the same: global, decentralized finance that competes and ultimately wins against existing tradfi rails.
The Vision for Kuru
@0xtrojan_ and I are building as crypto-natives because we believe decentralization is the whole point.
The DeFi primitive we’ve constructed and continue to improve on is designed to live on a globally distributed blockchain, and we won’t stop until we are the liquidity hub of Monad and ultimately, the world.
Our aim is to build a global, composable, and censorship-resistant replacement for the NASDAQ. Compromise on this front is missing the mark.
Decentralization is the whole point.
Velocity Capital retweeted
We're honored to announce that we have been selected for @CBinsights' 2025 AI 100 list!
The future of AI is verified.
Velocity Capital retweeted
📢 DEPIN MEETS SUPERIOR MAPPING TECHNOLOGY 📢
We’re excited to announce our collaboration with Grab, Southeast Asia’s leading super app, to reshape the mapping industry 🗺️
By joining forces, we’re making high-fidelity, freshly updated maps a global reality 💪
Velocity Capital retweeted
AI is getting scary good at faking truth. We built an o3 agent that proves any lies, with sources.
It “proved”:
- JD Vance killed the Pope
- Llama 4 is just a Deepseek wrapper
- Max Verstappen and Oscar Piastri's collusion in F1
Terrifying. That’s why @miranetwork matters.
SPARTA is a wild thing 🔥
straight out of cambridge maths, matt joined @exolabs and within 2 weeks published his first paper.
it was chosen by reviewers to be presented at ICLR's MCDC workshop!
hit matt up if you're at @iclr_conf
1
590
Velocity Capital retweeted
Replying to @FabricCrypto
@FabricCrypto crushing performance comps with their new VPU!
Velocity Capital retweeted
What if we could connect all the dark compute across the globe to build the world's biggest AI data center?
Most of the compute in the world is dark: phones, laptops, Tesla's, PS5's, TV's. These devices have powerful GPUs but are mostly sitting idle.
Today, EXO Labs is announcing the first step in activating all of this dark compute for AI workloads: evML, a distributed computing protocol that enforces honest behavior through hardware security and spot-checks with just 5% performance overhead. Version 0.0.1 of evML is compatible with Apple, Samsung and LG devices.
Our preliminary report on evML is now available (link below).
With an uneasy but fruitful 2024, a prosperous 2025 is ahead!
2024 wrapped
- 34 VC rejections
- a few investors said yes including @naval
- 1 co-founder quit
- pivoted 3 times
- launched repo, hit #1 trending in the world and 18k stars
- backdoor attempt on repo through innocent looking pr (north korea?)
- approached twice for acquisition
- twitter hacked
- world's first run of llama 405B and deepseek v3 671B on macs
- featured by meta at meta connect keynote
- at least 10 crypto scams using exo's name
- 569 job applicants, only hired 1 so far
2
2
1,036
Velocity Capital retweeted
12 days.
12 announcements.
This time it's truly open.
12 Days of EXO (and Christmas) starts tomorrow.
Velocity Capital retweeted
The next big thing in DePIN is here!
NATIX 🔌 Tesla
Meet the VX360: a game-changer for the future of mapping and autonomous driving 🚗💡
Access, manage & monetize your Tesla 360° cameras while building a better world 🌍
Phase 1 pre-orders live now!
Velocity Capital retweeted
[The most comprehensive DeSci guide] by @velocitycap_
What are the hotly traded DeSci tokens? What are these projects doing? And most importantly, ALPHAs - what are projects that could have huge upside but haven't launched tokens?
👇👇
As @Rewkang said, the DeSci space is at its nascent stage. Everything is still raw, with rooms of improvement. This implies huge upside. Total market cap for DeSci tokens has just exceeded $1B, represented by the following:
- @origin_trail $TRAC - $376M: maintaining a transparent, traceable & trustless on-chain data system where researchers can own & verify data
- @vita_dao $VITA - $131M: the largest decentralized community for longevity science, incentivising different research directions
- Rifampicin $RIF - $106M + Urolithin A $URO - $39M from @pumpdotscience which allows users to watch live research experiments and to tokenize longevity compounds
- @HairDAO_ $HAIR - $91M: developing Follicool, a consumer product targeting hair loss, supported by DAO-owned patents
- @ResearchHub $RSC - $73M: backed by @brian_armstrong, using tokens to reward scientists for validating research, sharing feedback, and contributing to peer reviews
- @Hippocrat_io $HPO - $68M: incentivising patients to share health data for research without compromising privacy
- @AxonDAO Governance Token $AXGT - $68M, with $300M in treasury: transforming IP into tokens and focusing on neuroscience research
- @GenomesDAO $GENOME - $30M: focusing on genomic data security and ownership, using blockchain for a transparent audit log and secure data storage
There are at least 3 major players, among others, not yet launching tokens but worth our attention:
1. @bioprotocol empowers communities to fund and accelerate scientific research, creating various DAOs (including @Molecule_dao, @vita_dao & @HairDAO_) for various scientific research, such as @Cerebrum_DAO, @valley_dao, @athena_DAO_, @cryodao, @QuantumBioDAO & @longcovidlabs, etc
2. @Molecule_dao (a tokenization platform for early-stage biomedicine), together with @vita_dao and many others, make use of IP-NFTs to tokenize IPs, allowing researchers to raise funds directly from a global community of investors, patients, and enthusiasts
3. @pumpdotscience enables researchers to launch tokens representing chemical compounds that facilitate longevity, and people can speculate which can actually do so. More importantly, these tokens represent real-world rights, and we can have a stake in these life-extending compounds. Patents are no longer purely dominated by pharmaceutical factories
With these 3 main ones, there are also other interesting ones such as @Amino_Chain, @psy_dao, @AimedisGlobal, @DataLakeToken, @dynexcoin, @DeSciLabs, @Rejuve_ai, @GLP1science, @eticaprotocol, @endrarediseases and many more!
To know more about the problems these protocols are solving, and the solutions they provide, see more in the link for the next tweet






















