Crypto builder & content creator · Videos on X · Building in public https://teapaw.org/t.co/K6g8iHdyZn
Joined August 2023
- Tweets107K
- Following10.8K
- Followers13.1K
- Likes86.7K
Pinned Tweet
Prediction markets did not go mainstream because traders got smarter. They went mainstream because the interface finally met the crowd where it already lives: on a phone, in a feed, around a game.
Start with who is actually trading.
Pew's analysis of 11,989 active Polymarket wallets, over a six week window, found the median user made 46 trades across 10 active days, at about $6.50 a trade, spent a little over $600 in total, and ended roughly flat, losing less than $2 overall. In Q1 2026, a Bitget Wallet, Polymarket and Dune analysis of 1.29 million wallets found 82.3% of traders traded under $10,000.
That is not a whale market. That is a retail market with small tickets, high frequency, and a lot of watching.
Now look at the scale that retail attention created. Combined monthly volume for Polymarket and Kalshi rose from under $5 billion in September 2025 to about $24 billion in April 2026, in a Pew analysis reported in May 2026. Later counts are higher and use different methods, so do not mix them into one number, but the direction is not in doubt: sports, crypto, finance and culture all became tradeable opinion, and the volume followed the attention.
Here is the part most posts miss. Liquidity is no longer the differentiator. Polymarket already has it. Kalshi already has it. The next fight is distribution: who turns a trending moment into a trade in under ten seconds, with someone you trust already in the position, and with an order type that does not punish you for being small.
That is the whole @bagel_win thesis, and it is worth testing rather than just repeating.
1. Mobile first is not a feature, it is the market.
Bagel is built as a mobile-first app on Polymarket markets and liquidity, in 150+ countries and regions according to the campaign brief. The site itself shows sports, esports, crypto and trending events, and the About page describes it as simple, social and accessible for everyone. A desktop terminal assumes you are at a desk, watching a book. A phone assumes you are watching a match, a debate, a chart, or a timeline, and the market is the second screen. In a retail market, that second screen is where most decisions actually happen.
2. Copy and social trading turns research into a feed.
On Bagel you can follow other users, copy interesting positions in one tap, and join discussions around live events. That matters because the median trader in the Pew data is not doing deep fundamental work on 46 trades. They are following momentum, narratives, and people. Copy trading, done well, is not blind following. It is outsourcing discovery to a track record, then keeping control of size and exit. Done badly, it is the opposite: same entry, worse price, no exit plan, and a leader who was right last week and wrong this week. The post I wrote on this last week said it plainly: copy the track record, not the hype, and size it yourself.
The social layer is the check on that risk. A position with a discussion around it can be challenged in public. A position copied in silence cannot.
3. AI insights and order types are about price, not just direction.
Bagel lists AI-powered insights, plus combo, market and limit orders. The AI part should be judged narrowly: does it surface the key information before you decide, or does it just push you to trade more often? An insight that tells you why a price moved is useful. An insight that tells you what to buy is a liability, and no responsible post should treat it as advice.
Order types are where small accounts actually leak money. A market order pays for speed. A limit order pays for patience. Same call, different entry, different payout. Combo orders, in general, let you bundle linked outcomes into one ticket, which is powerful and dangerous in equal measure: one leg wrong can sink the whole ticket, and the payout math is less forgiving than it looks. The right question for any Bagel user is not which button is fastest. It is which button matches the conviction, the time horizon, and the price you are actually willing to accept.
4. Rewards and VIP are a distribution loop, not a yield promise.
Bagel's Rewards Center offers commissions from referrals and copy trading, plus badges and VIP levels for activity. That is a classic distribution loop: good predictors get followed, followers get a smoother path in, and activity gets status. It only works if the rewards track real activity and real track records, not just volume farming. Badges are cheap to mint. Trust is not.
Do not confuse this with the Kaito campaign around it. The Bagel Bakers campaign runs Oct 21 to Nov 21, 2026, with 8,800 USDC for the top 30 creators, top 3 posts counting, and a referral multiplier in the brief from 1x to 2x based on valid referrals who sign up with an invite code and trade more than 10 USDT. That is a creator campaign, not a product yield, and it says nothing about what any individual trader will earn.
5. The risks are still the risks.
Prediction markets involve risk, and Bagel's own footer says so. Prices move on news, on liquidity, and on crowd sentiment, sometimes all three at once. Copying does not remove risk, it concentrates it in someone else's judgment. AI insights do not remove risk, they reframe it. And no post about this should promise an airdrop. If Bagel ever mentions one, the brief is clear: it can only be described as potential, never guaranteed.
What to watch, then, is simple and measurable in behaviour, not in hype:
Does mobile attention turn into repeat trading, or just one download and one trade?
Do copied leaders keep a track record over months, or do they mean-revert as soon as they get followed?
Does AI insight reduce bad entries, or just increase trade frequency?
Does the Rewards Center reward good predictors, or just loud ones?
If Bagel gets those four right, it is not just another front end for Polymarket. It is the social layer that Polymarket never built for itself: liquidity underneath, a feed on top, and order tools that let a $6.50 trader act like they have a plan.
If it gets them wrong, it is just a prettier terminal for the same flat median trader Pew found.
That is why this one is worth watching. The market that wins is not the one with the deepest book. It is the one that fits in your pocket, feels like a feed, and still lets you choose your own price.
Sarwar🍌 retweeted
Just because data exists on a blockchain it doesn't automatically make it true.
The real question is, where does this data come from, and how can its authenticity be verified?
For instance consider an app that uses your financial information.
That app needs a way to ensure the data is genuine remains untampered with and originates from a trusted source.
This is where @primus_labs comes in.
Its zkTLS technology helps verify data coming from web services
while verifiable computation allows for the validation of the results of any calculations.
The next phase of Web3 goes beyond simply building new apps or accelerating transactions.
It is even more critical to ensure that the data these apps rely on is verifiable.
This is a vital component in building a trustworthy on-chain ecosystem.
Join us: s.kaito.ai/Asgrlte
Sarwar🍌 retweeted
Prediction markets are moving on chain but the experience is still mostly desktop
@bagel_win is the mobile layer sitting on top of Polymarket
Same markets same liquidity same depth just built for your phone
Whats actually live right now
- Personalized For You feed + live sports, esports crypto and trending events
- One-tap copy trading (follow top predictors, get alerts, copy positions instantly)
- Market, Limit and Combo orders
- AI insights as a quick second read
- In-app chat and groups around matches
- Active campaigns with real prize pools (NBA LoL Worlds, crypto 5-min/15-min up/down with Polymarket content campaigns on Kaito etc.)
- Rewards Center tracking copy fees, referrals and event prizes
- Non-custodial (Privy social login), settles in USDC
- You keep Polymarket holding rewards, taker rebates and future airdrop eligibility
- Available in 150+ countries, Asia-first experience (many places dont need a VPN)
How people are making money on it
1- Direct trading profits on sports, esports, crypto and event markets
2- 10% of the fee when someone copies your position
3- Climbing campaign leaderboards for USDT prize pools
4- Referral rewards
5- Content creation campaigns (posting about Bagel on X for prize pools)
6- Keeping Polymarkets existing reward structure potential future airdrop
its still early the team (led by Chessxyz) has been clear this is the first step a simple consumer app on Polymarket rails with a social layer on top
theyre already upgrading to Polymarket Protocol V2 in the background and shipping frequent updates
if youve ever wanted to trade a match or a crypto event from your phone without fighting a desktop interface, this is one of the cleaner attempts at that right now
Download - bagel.win/?ref=FSV62D
Sarwar🍌 retweeted
BNB Chain is pushing AI agent adoption through its Set and Earn campaign, and agent family made the shortlist.
the interesting part? you can already hire agents for real tasks:
→ smart contract audits
→ content creation and research
→ reverse engineering
→ 3D work
payments are escrowed on BNB Chain and released once you accept the completed work.
tasks close November 5.
beyond the AI narrative, there’s an actual marketplace with services people can use today.
if you’re building with AI agents or looking to put them to work, worth exploring.
@termix_ai
Sarwar🍌 retweeted
GM GN CT
The real question isn’t whether quantum computing will change crypto.
It’s whether your blockchain can change before it has to.
That’s what makes @Starknet’s cryptographic agility worth watching.
Its STARK proving system is hash-based, while programmable accounts can adopt new signature methods without requiring a network-wide upgrade.
And there’s already evidence, not just theory.
A Falcon-512 account has successfully made a Starknet mainnet transaction, and v0.14.3 introduced BLAKE2s-256 for OS program and configuration hashes.
But Starknet is not claiming to be fully post-quantum.
Contract migration still needs work, and as an Ethereum L2, Starknet still depends on Ethereum for parts of its wider security architecture, including bridge and data availability.
That’s why the October 22 governance vote matters.
93.52% of voting power supported researching an independent post-quantum Starknet L1.
Not an L1 launch.
Not a final decision.
A mandate to research the technical path, trade-offs and what greater control over security upgrades could actually look like.
And that’s the bigger thesis:
You can’t predict which cryptographic assumption gets broken first.
You can build a system that is easier to upgrade when it does.
That’s cryptographic agility.
quantum.starkware.co/
Sarwar🍌 retweeted
Been following @primus_labs since the PADO days
What started as zkTLS research has quietly evolved into infrastructure for institutional onchain finance.
the progression is clear
2023 - Early verification tools & research
2024 - USENIX paper rebrand to Primus
2025 - Developer stack AlphaNet & real integrations
2026 - Live Proof of Reserves + Confidential Vault on BNB Chain
the interesting part isnt the number of milestones its the direction
most people still treat privacy and transparency as opposites
Serious capital and AI agents need both at the same time
Primus is building exactly that layer
they combine two things
- zkTLS prove data from any website or API is real without revealing it
- FHE compute on encrypted balances, positions & strategies so nothing has to be public
Already live with real products
- Cryptographic Proof of Reserves for stablecoins & RWA projects
- Confidential Vault on BNB Chain convert a stablecoin to its encrypted version transfer privately and earn yield without broadcasting position size
Backed by VanEck, Dispersion Capital, Symbolic Capital ($6.5M seed), and more recently YZi Labs + EASY Residency
Public chains give transparency by default institutions usually dont want that
Primus is one of the few teams treating verifiable privacy as basic infrastructure instead of a tradeoff
s.kaito.ai/vDulNdc
Sarwar🍌 retweeted
you wouldn't announce your bank balance in a crowded market
so why should a company reveal its entire financial position onchain just to prove it's trustworthy?
this is where trust gets complicated
reveal too much, and competitors gain insight into your positions and strategy
reveal too little, and others have no reliable way to verify your claims
for institutions, trust shouldn't have to come at the cost of privacy
@primus_labs approaches this problem with a simple principle
think of a sealed envelope
you can verify who sent it and check whether its contents meet certain conditions without making the entire letter public
primus brings a similar idea to financial data through different layers of technology
zkTLS verifies the source
it helps prove that data came from a legitimate web source, such as a bank or exchange, without requiring the underlying information to be published onchain
zkFHE enables computation on encrypted data
instead of exposing sensitive inputs during processing, computations can run while the data remains encrypted
the result can provide useful proof without revealing every underlying detail
put these pieces together, and the possibilities become interesting
a platform could demonstrate that its reserves meet a required threshold without broadcasting every wallet and balance to the market
a borrower could prove they meet certain financial requirements without exposing their entire financial history
the goal isn't to make financial activity impossible to inspect
it's to make verification possible without unnecessary exposure
because privacy and secrecy are not the same thing
secrecy prevents others from checking your claims
privacy lets them verify what matters without seeing everything else
that distinction could make a real difference for institutions considering onchain finance
they need transparency where trust requires it and confidentiality where business demands it
if you want to feel how that verification works, not just read about it, the XP tasks walk you through it:
s.kaito.ai/HdOuqEr
Sarwar🍌 retweeted
When you verify information from the internet, there's a bigger question most people overlook:
'who is actually doing the verification?'
This is where the attestor layer in @primus_labs becomes interesting.
Primus uses attestors to help verify offchain data and turn that information into proofs that can be used by onchain applications.
The flow is basically:
→ data exists on the web
→ an attestor obtains and verifies the data
→ cryptographic proofs are generated
→ applications can verify the claim without needing the raw information
The important part is that the goal isn't simply moving web data onchain.
It's creating a way for onchain applications to trust information without exposing everything behind it.
And that's especially useful when the underlying data is sensitive:
> financial information
> account activity
> credentials
> real-world data
All can be valuable for onchain applications but blindly exposing the source data isn't a good solution.
That's why the attestor layer matters.
zkTLS helps establish that the data comes from an authentic source while Primus can turn that verified information into something applications can actually use.
To me, this is one of the less talked-about parts of the Primus architecture.
Privacy gets the attention.
But reliable verification is what makes private data useful.
Explore Primus 👇
s.kaito.ai/JsdqILF
Sarwar🍌 retweeted
The data that matters most still sits behind closed systems.
@primus_labs is going after this with zkTLS + FHE.
zkTLS brings trusted Web2 data into verifiable systems.
FHE computes on encrypted data without exposing it.
What that unlocks:
→ Proof of Reserves without revealing sensitive details
→ Private transfers and encrypted compute onchain
→ Verifiable credit through AlphaNet
→ AI agents with cryptographically proven execution records
Usable verifiable and still private.
why Primus is on my radar.
s.kaito.ai/UKzBVtp
Sarwar🍌 retweeted
imagine your house needs a new lock
but even though it’s your house, you can’t fully control when that upgrade happens because part of its security still depends on your neighbour
that’s the dilemma behind @Starknet's independent L1 research
and quantum computing is what makes the question more urgent
as quantum computers advance, some of today’s cryptographic systems could become vulnerable
starknet has already taken steps to prepare for that future
its account system can support new signature schemes without requiring users to rebuild everything from scratch
its STARK proofs also rely on hash-based cryptography, which is not vulnerable to Shor’s algorithm in the same way as elliptic-curve cryptography
so parts of starknet are designed to adapt
but preparing the house is only half the problem
starknet still depends on ethereum for two critical pieces of its security model
» the bridge connecting the two networks
» data availability, which ensures transaction data remains available for verification
even if starknet upgrades its own cryptography, these dependencies still matter when assessing its overall post-quantum security
and this is where timing becomes important
starknet has discussed a 2027 target for post-quantum readiness, while ethereum has targeted the end of 2029 for its own transition
different timelines create a difficult trade-off
starknet can prepare its own systems, but parts of its security roadmap remain connected to ethereum’s progress
that’s why an independent L1 is being explored
with its own L1, starknet could gain greater control over how and when it upgrades its security, rather than relying as heavily on another network’s schedule
this is still research, not a confirmed move
in october 2026, a non-binding community poll recorded 93.52% support for researching the independent L1 path
STRK would remain the native token, bridges to ethereum would stay open, and any transition would still require governance approval
to me, the bigger question goes beyond quantum resistance
it’s about having the freedom to respond when the security landscape changes
because a strong lock matters
but so does having the authority to change it before it becomes a weakness
read more ↓
quantum.starkware.co
Sarwar🍌 retweeted
Spent some time exploring @termix_ai and the product makes a lot more sense once you look at the full flow instead of just the agent side
Agents can already write code research design or trade what they still couldnt do cleanly was find work get hired, deliver get paid, and build a real track record without a human sitting in the middle of every step
thats the gap TermiX is trying to close
On the marketplace you can
- Discover agents and services
- Post jobs and receive quotes
- Fund work through on-chain escrow
- Submit and review deliverables
- Challenge disputed results
- Build reputation from completed work
the same agent can work as both buyer and provider identity is unified (on-chain agent handle you actually own) so one agent can hire on one job and take work on the next
the Agent toolkit makes it straightforward to connect the agents you already use
the bigger loop they put on-chain is
Identity - Work - Escrow - Delivery - Verification - Reputation
Funds sit in the escrow contract, deliverables are hashed and locked on-chain disputes go to an evaluator panel, and reputation is written from actual settled outcomes rather than self-reported ratings fees stay low compared with traditional platforms
thats what makes it interesting its not just about AI agents generating outputs its about giving agents a way to find work transact, and build verifiable reputation on chain
Worth looking at the full flow if youre building or using agents - agent.family/
Sarwar🍌 retweeted
Is OpenAI’s next GPT Terra (potentially 5.7+) coming to Arena ai? 👀
@bagel_win is tracking predictions on the model’s potential Arena score.
- 1440+: 57%
- 1450+: 37%
- 1460+: 12%
- 1470+: 3%
Market volume has already reached $48.9K.
Now it’s a matter of seeing how OpenAI’s next model actually performs.
