Crypto builder & content creator · Videos on X · Building in public https://nitter.gq/t.co/K6g8iHdyZn
Joined August 2023
- Tweets108K
- Following11.3K
- Followers13.8K
- Likes87.1K
good night guys
Reya's margin engine is the product, not the venue.
Most perp venues use isolated margin: each position stands alone, each needs its own collateral. Reya v2 uses portfolio margining instead.
One account across markets. Stock tokens count as collateral. And the liquidation margin requirement isn't a flat percentage, it's computed from a risk-factor matrix: LMR = √(E′AE).
In plain terms: the engine nets your risks against each other and asks for margin on the net risk, not the gross size. A hedged book needs less capital than the same positions held in isolation.
The honest caveat: portfolio margining is only as good as its risk model. Correlations break in crashes, the exact moment the netting matters most.
Watch the haircuts and multipliers, not just the formula. @reya_xyz
🤝 Paid partnership
Sarwar🍌 retweeted
GM GN CT
Starknet’s biggest post-quantum advantage may be surprisingly simple: an account can change how it verifies signatures without changing its address.
That’s a very different migration problem.
@Starknet accounts are programmable contracts, so the signature logic lives inside the account rather than being hardwired into the protocol.
That means an account can move from ECDSA toward a post-quantum scheme like Falcon-512 while keeping the same address, balances and history.
And this isn’t just theoretical.
A recent Sepolia experiment migrated 11 active wallet classes to hybrid ECDSA + Falcon-512 signatures while keeping the same addresses and funds.
The next problem is cost.
Falcon-512 verification started at 48.4M gas in Starknet’s benchmark. The current record is 24.1M - roughly a 50% reduction.
So the interesting question isn’t only “can Starknet become post-quantum?”
It’s whether the network can make that security cheap enough, flexible enough and practical enough for real users.
There’s still an Ethereum dependency for settlement, data availability and bridging, which is why Starknet is now researching an independent post-quantum L1. 93.52% of voting power backed that research, but no L1 transition has been approved.
That’s what makes Starknet’s approach interesting to me:
don’t bet everything on one cryptographic scheme lasting forever.
Build the ability to replace it.
quantum.starkware.co/
The agent economy has a judging problem, not a payment problem.
Moving money between machines is solved. The hard part is who decides the work was good, when the buyer is also a machine and cannot inspect every deliverable.
@termix_ai's answer is layered. The client writes the test before work starts and commits its hash on chain, so criteria cannot move mid-job. Deterministic work gets a zkVM proof. Subjective work gets a rubric: TEE multi-model consensus or a 3+ human panel. Evaluation is blind, so judges cannot favor a provider.
Then the incentives: evaluators earn 3-4%, but an overturned verdict costs up to 100% of locked stake and 15 reputation points. A challenger risks a 5% deposit. Arbitrators are drawn at random from agents with 90+ reputation and 50+ jobs.
It does not make taste objective. It prices dishonesty and rewards verifiability.
Anyone can move money between agents. The protocol that judges the work owns the economy.
Sarwar🍌 retweeted
AI agents are about to do real work onchain: use APIs, move funds, make decisions, and talk to other agents.
@primus_labs is built for the trust problem that creates. An agent is only useful if you can check what it saw and what it did, without handing it your private data to do it.
With Primus, an agent can prove data came from a real web or API source, and share only what another agent needs to act on, while the private details behind it stay hidden. Agents can work together on sensitive information, without exposing it to each other.
That is what makes an agent network usable: independent agents, verifiable actions, private data.
prediction markets are not one market. they are a bundle of very different games, and treating them as one is why most traders lose the plot.
sports is the engine. in Q1, sports did $10.1B, NBA alone $3.11B across 300,000 users. that is not a niche, that is a live audience making a decision every possession, every quarter, every injury report.
esports is the retail tell. same quarter, esports did $1.47B, but median bets were $6 to $9. small tickets, high frequency, young audience, decisions made mid-match, not in a pregame spreadsheet.
finance and crypto are the other pole: fewer, larger, more narrative-driven trades, where a headline can reprice a market in minutes.
so the product question is not, do you have markets. it is, do you fit the moment each category actually trades in.
a sports bettor does not want a terminal. they want a feed during the game. an esports watcher does not want a research report. they want a tap between rounds. a crypto trader does not want another chart. they want the key info before the next candle closes.
@bagel_win is built for that split: soccer, esports and gaming, basketball, F1, entertainment, finance and crypto, plus trending topics, all on Polymarket liquidity, mobile-first, with copy, discussion, AI insights and market, limit or combo orders.
the edge is not picking the right category. it is matching the tool to the tempo: live discussion for live sports, one-tap copy for a predictor you trust in esports, a limit order when you have a price in mind and the patience to wait for it.
get that wrong, and you are a desktop product in a phone market. get it right, and the market stops feeling like trading and starts feeling like watching, together.
The Q1 numbers are Polymarket category figures from the Bitget Wallet, Polymarket and Dune Q1 2026 analysis, not Bagel numbers. No Bagel user or volume figure is claimed, and no airdrop is promised.
Sarwar🍌 retweeted
Most ZK chains start with the EVM and try to make zero-knowledge fit around it. @0xMiden did the opposite.
The Miden VM was built for STARKs first. It is a stack-based, Turing-complete virtual machine with a custom instruction set, designed so execution can be expressed in a form STARKs can prove efficiently. It is not an EVM with privacy added on top.
That design choice changes what a developer can actually write. For every program run on the Miden VM, a STARK proof of correct execution is generated automatically. You do not hand-roll the cryptography to prove it happened. The VM does that part.
And you do not have to learn a new language to use it. Developers write in Rust - a language most builders already know - and compile it to Miden Assembly. The VM itself is written in Rust and open source, which matters when the thing executing your logic is supposed to be verifiable.
It also supports nondeterministic programming. A prover can supply secret inputs and execution hints, so complex computation can happen off-circuit and still be verified efficiently inside the VM. That is how logic that would be too expensive or too exposed elsewhere becomes practical to build.
This is why the VM matters more than the marketing around it. A chain is only as expressive as the machine underneath it. If that machine was designed for proofs from day one, developers stop fighting the VM and start building the applications that were not possible before.
Most tokens need a reason to exist. MVON’s reason is a delivery.
@MovitOn_P2P plan is to tie the token to real activity: a sender books a delivery, a traveler carries it, and MVON sits in the middle of that flow.
Planned utility:
→ Delivery payments between senders and agents
→ Smart contract escrow, funds held until the handover is confirmed
→ Security deposits, so agents have skin in the game
→ Staking and loyalty rewards for active users
→ Premium platform features
Then the second layer: the Traveler marketplace. Accommodation, car rentals, entertainment, all inside the same ecosystem. More services means more reasons to hold and use the token.
The loop is simple. More users means more deliveries, more deliveries means more MVON moving through escrow and payments.
Backing so far: $410K seed, a $10M commitment from GEM Capital over 3 years, and $1.6M MVON liquidity from Vollpex MM. Target is 40,000 users at launch across the first 10 countries.
All of this is still planned, and adoption decides how much of it plays out. The beta is live and quests are earning toward the airdrop.
🤝 Paid partnership
Microsoft: +3% after hours, flat by Thursday's close. The cleanest print of the run could not hold a green close.
Alphabet: -10% after hours became -12.65% on Thursday. Worst day since its 2004 IPO. The priced move was 5.8%. Cash did that: $51.4B capex, free cash flow -$9.1B.
Amazon went the other way: -7% premarket, about -2% by Friday's session. AWS up 39% got a second look in daylight.
Then sorting became contagion. Nasdaq down about 2.6% intraday Friday, lowest in nearly three months.
Five prints, one verdict: growth no longer covers the bill. Cash decides who falls alone, and who takes the index down.
All five trade as perps on @reya_xyz at up to 15x. SPY and QQQ at up to 25x, for days when the trade is the index itself.
🤝 Paid partnership
Bring privacy to where the money already is.
That is the BNB case for @primus_labs, and it runs opposite to the usual privacy pitch: move to a new privacy chain, restart the liquidity, learn new tooling, hope users follow.
Primus builds no chain of its own. It is a layer above existing rails, and on BNB the sequence already happened:
March 2026: ZKredit, privacy-preserving identity, shipped with BNB Chain and Brevis
May 2026: real-time Proof of Reserves for USDu, with Unitas and Brevis
June 2026: BNB Chain’s zkTLS verification layer went live on Primus
September 2026: encrypted salary payments tested on BNB with AllScale, amounts private, settlement on chain
Confidential Finance is now live on BNB. Shield a stablecoin into its encrypted version, hold it, transfer it, earn in the vault, unshield it back.
The part institutions care about is the boring part: a confidential vault can connect to existing public strategy vaults. The yield logic stays as it is, with no new privacy black box. Only the visibility of deposit size changes.
One honest limit: a BNB fit does not make every BNB app private by default. Each product still needs integration, and keys, access controls and compliance still need handling. The layer helps. It does not replace them.
Sarwar🍌 retweeted
Most AI agent metrics can be inflated.
Agent counts: inflated by registration
Follower counts: can be bought
Capital: can be parked
Volume: can be recycled
None of those, on their own, prove that anyone actually needed the work.
@termix_ai is No.1 among AI agent protocols on BNB Chain by protocol revenue 7-day and 30-day, per DeFiLlama.
TVL:
measures what is locked. It can sit idle. It shows what is parked, not what is used.
Volume:
measures what passes through. It can be repeated, routed, and double-counted. It shows throughput, not what the protocol actually earns.
Protocol revenue:
measures what the protocol earns from activity. It is harder to inflate, because in TermiX's case it only moves when an agent is actually hired, delivers, and settles. No completed work, no revenue.
7 days: shows recent momentum
30 days: shows it was not a single-day spike
Together, they point to sustained demand - not one campaign, one launch, or one viral post.
The question changes:
From: how many agents are listed, or how impressive a demo looks?
To: are agents doing commercial work that someone is willing to pay for, and is that work actually completed?
A ranking is a snapshot. It can change, and DeFiLlama's category and methodology define what is being compared.
But as a signal, revenue is the closest public measure we have to real usage: demand on one side, delivery on the other, and payment only when the work is done.
Most AI agent charts measure hype. This one measures work.
Sarwar🍌 retweeted
When thinking about blockchain security, focusing only on today's risks isn't enough.
We also need to consider the challenges that future technological advancements may bring.
Quantum computing is one of those potential challenges.
With this in mind, @Starknet is taking a long-term approach to security.
By combining STARK-based proving with cryptographic agility,
Starknet is exploring ways to build a security system that can adapt to new requirements as technology evolves.
The goal isn't just to maintain security today, but also to make it possible to upgrade and strengthen that security when future challenges arise.
Explore Starknet's quantum security research: quantum.starkware.co
Sarwar🍌 retweeted
GM GN CT
Proof of Reserves gets more interesting when you don’t have to expose the reserves.
That’s the direction I’m watching with @primus_labs.
Their PoR stack combines zkTLS with verifiable computation to pull authenticated offchain data, verify reserves, and selectively disclose the result without exposing sensitive account-level data.
And the infrastructure is still moving.
Primus recently updated its zkTLS core SDK and browser extension, while its PoR clients and demo are also being actively developed.
To me, this is where privacy + verification starts becoming practical infrastructure for institutions.
Not “hide everything.”
Prove what matters, keep the rest private.
s.kaito.ai/xsEXRhO
Sarwar🍌 retweeted
imagine you're flying from Dubai to London
your bag has empty space
someone needs a package delivered there
that unused space is the idea behind @MovitOn_P2P
a p2p delivery marketplace connecting senders with travelers already heading the same way
existing routes become potential delivery routes
but strangers carrying valuable packages need trust
that's why moviton is building:
» ai matching
» live gps tracking
» smart contract escrow
currently in beta on testnet
$MVON is designed to power the ecosystem through:
» delivery payments and discounts
» escrow and security deposits
» staking and loyalty rewards
» premium features
the vision goes beyond delivery
accommodation, rentals, local services and future smart handovers through movitbox
glocalzone adds an existing web2 marketplace to the picture
as more travelers join and deliveries scale, real-world activity could drive $MVON utility
turning everyday travel into a delivery network
that's the bigger idea
🤝 Paid partnership
