Web3 Analyst | BD | RWA • AI • DeFi | Decoding digital asset narratives.
Web3
Joined October 2020
- Tweets47.4K
- Following2.5K
- Followers4.8K
- Likes128K
I’ve watched the Tesla Roadster demos so many times, and I still can’t fully wrap my head around what this car is supposed to do.
The idea of using suction to increase grip, combined with adaptive aerodynamics and cold gas thrusters, sounds almost unreal.
But I’m more interested in seeing how much of this actually makes it into the production version.
The engineering possibilities are fascinating, but the real test is whether Tesla can deliver what it’s promising.
If they pull this off, it could change how we think about performance cars.
Never bet against Elon Musk.
Zevryn retweeted
Crypto is one of the few industries where you can do almost everything right and still lose everything.
You can spend years learning the market, researching projects, managing risk, and building a portfolio, only for one exploit to wipe out years of progress.
You can build your reputation on X for years, create valuable content, grow a community, and wake up to a suspended account.
You can trust a platform to safeguard your assets and discover that your security depends on a company you barely know.
You can hold a solid project, support it through every market condition, and watch attention and liquidity move elsewhere overnight.
And the worst part? There isn’t always someone to hold accountable, and there isn’t always a way to recover what you’ve lost.
Crypto rewards independence, but that independence comes with a level of personal responsibility most people underestimate.
Your money, security, reputation, and sometimes even your livelihood can depend on systems you don’t fully control.
That’s why surviving in crypto requires more than being smart or finding the next big opportunity.
It requires risk management, security, diversification, adaptability, and the ability to start again when things go wrong.
The goal isn’t just to make it in crypto. It’s to build something that can survive crypto.
Crypto is one of the few industries where you can do almost everything right and still lose everything.
You can spend years learning the market, researching projects, managing risk, and building a portfolio, only for one exploit to wipe out years of progress.
You can build your reputation on X for years, create valuable content, grow a community, and wake up to a suspended account.
You can trust a platform to safeguard your assets and discover that your security depends on a company you barely know.
You can hold a solid project, support it through every market condition, and watch attention and liquidity move elsewhere overnight.
And the worst part? There isn’t always someone to hold accountable, and there isn’t always a way to recover what you’ve lost.
Crypto rewards independence, but that independence comes with a level of personal responsibility most people underestimate.
Your money, security, reputation, and sometimes even your livelihood can depend on systems you don’t fully control.
That’s why surviving in crypto requires more than being smart or finding the next big opportunity.
It requires risk management, security, diversification, adaptability, and the ability to start again when things go wrong.
The goal isn’t just to make it in crypto. It’s to build something that can survive crypto.
Recursive Self-Improvement Could Become AI’s Most Important Breakthrough
One of the most significant developments in AI is the growing use of AI systems to accelerate the development of more capable AI systems.
Known as recursive self-improvement (RSI), the concept involves AI contributing to the processes that improve its own capabilities, including software development, experimentation, evaluation, and research.
As of October 2026, interest in this topic is reportedly rising sharply, with approximately 135,000 monthly searches and 20× year-over-year growth.
The more important signal, however, is the progress in practical applications.
AI is increasingly being integrated into software engineering and research workflows. @AnthropicAI has reported substantial AI involvement in its codebase
while @OpenAI and Google DeepMind are pursuing systems that automate parts of coding, experimentation, and research.
These developments represent meaningful progress, but they do not yet demonstrate fully autonomous, open-ended recursive self-improvement.
Human oversight, evaluation reliability, computing requirements, and research constraints remain significant limitations.
Other emerging trends reinforce the broader shift toward more capable, action-oriented AI:
➯ Project Glasswing: Applying frontier AI to cybersecurity and vulnerability discovery.
➯ Microduck: Bringing open-source robotics and embodied AI to developers through an affordable bipedal platform.
➯ Codex Micro: Exploring dedicated hardware for monitoring and controlling AI coding agents.
➯ Pocket AI: Extending AI into wearable tools for transcription, summarization, and task management.
➯ AI-powered game modding: Making complex modifications more accessible through coding agents, while raising questions about quality and originality.
The common thread is the transition from AI that primarily generates responses to systems that can execute tasks, operate tools, support research, and interact with the physical world.
Recursive self-improvement stands out because it could influence the rate at which these other capabilities advance.
If AI systems can reliably accelerate the research and engineering required to build better models, improvements in AI development could become increasingly significant.
However, the scale and speed of that effect remain uncertain.
The key metric to watch is not simply how much code AI writes or how many tasks it completes.
It is whether AI can measurably shorten the cycle of research, experimentation, evaluation, and deployment while producing consistently better results.
That is where the distinction between incremental automation and genuine self-improvement will become clear.
Zevryn retweeted
I studied 7 tokenized US stocks on @coingecko’s RWA list, and one thing stands out: price tracking is tight, but spot volume and perps activity tell two different stories.
$NVDAc leads with $160M in tokenized mcap and $79M in spot volume, while its perps volume sits at $1.23B.
$GOOGLc records $40M in spot volume against $824M in perps. $AAPLc has $29M vs $770M, while $METAc records $39M vs $769M.
The price gaps are just 0.0–0.1% across all 7 assets in this snapshot.
But look at the volume split. Perps activity is significantly higher across every name, while tokenized spot volume varies from $9M on $MSFTc to $79M on $NVDAc.
That’s the part I wanted to study more closely. A tight price gap tells us what the displayed prices look like at that moment. It doesn’t tell us everything about spot liquidity, execution quality or redemption.
There’s also a wider RWA angle here. The list showed $9.426B in tokenized mcap, with gold alone accounting for roughly $6.01B. These seven tech stocks are only one part of the market.
One detail worth keeping in mind: CoinGecko’s RWA list and charts page showed different aggregate volume figures in the same time window. Always check which page and metric you’re using before drawing comparisons.
CoinGecko tracks this data; it doesn’t issue or custody these tokens.
I’ve attached the rows so you can compare the numbers yourself.
coingecko.com/en/rwa
Zevryn retweeted
One thing I pay attention to in crypto is how people can get more use out of the capital they already have.
You could hold NVIDIA shares with a traditional broker and still need separate capital to trade BTC. Your money is split between two systems.
@bitget rToken model offers another way to connect them.
Here’s how the stock transfer works:
1. Open Bitget and go to More → Stock Transfer.
2. Select your broker and the eligible stocks you want to move.
3. Get the receiving instructions provided through Alpaca Securities.
4. Request a Free of Payment (FOP) transfer from your existing broker.
5. Once the shares arrive and settle, Bitget’s In-Kind Tokenization Network automatically mints matching rTokens at a 1:1 ratio and credits your spot account.
No selling your shares and no separate manual conversion.
Bitget supports over 3,000 eligible U.S.-listed stocks and ETFs.
What matters to me is what happens after the transfer.
Eligible rTokens can be used as collateral in Bitget’s Unified Trading Account. For example, you could use rNVDA to support a BTC perpetual position without selling your NVIDIA exposure.
That's another way to put existing capital to work, although using stocks as collateral comes with liquidation risk.
The transfer-in fee on Bitget’s side is $0, but your broker may charge a fee. Processing typically takes around 10 business days.
Bitget has reported over $200M in rToken AUM and more than 4 million cumulative trades.
There’s also a $1M rSPCX rewards campaign, with rewards starting at $200 for qualifying transfers of $5,000 or more. Holding conditions apply.
To me, the real point is simple: tokenization becomes more useful when you can do more with an asset than just hold it.
Explore it yourself:
bitget.com/launchhub/stocks-…
Zevryn retweeted
Nvidia has one stock ticker, but its tokenized market is spread across multiple issuers, blockchains and trading venues.
I explored @coingecko's Nvidia page, and the interesting part is how these different markets connect.
The mid-October snapshot showed:
→ ~$159M in tokenized market cap
→ ~80.2M in 24H spot volume
→ ~$1.4B in 24H perpetuals volume
→ ~$785M in perpetuals open interest
But who is capturing that market?
Several issuers are building separate tokenized versions of NVDA.
➥ BTech Holdings ($NVDAB): ~$51M market cap
➥ xStocks ($NVDAX): ~$45M
➥ Ondo Finance ($NVDAON): ~$35M
➥ Robinhood Assets ($NVDA): ~$18M
➥ Coinbase ($NVDAC): ~$7M
➥ Reality ($RNVDA): ~$3M
Other issuers include Dinari, Backpack Securities and ST0x.
These aren't simply different tickers for the same product.
Each token has its own issuer, contract, blockchain, trading venues and liquidity profile. Their legal structures, redemption mechanisms and access restrictions can also differ.
Then there's the distribution layer.
The tracked markets span multiple chains and venues, including:
→ BNB Chain, Ethereum, Arbitrum, Base and Solana
→ Aerodrome, Uniswap, Raydium and PancakeSwap
→ Centralized exchanges such as Binance and Bitget
CoinGecko's markets section brings together hundreds of trading pairs, helping you investigate where activity is concentrated instead of assuming every version of NVDA trades in the same market.
And the derivatives numbers add another dimension.
With roughly $80.2M in tokenized spot volume versus $1.4B in perpetuals volume, derivatives activity is considerably larger in this snapshot.
But spot volume, derivatives volume and open interest measure different things. None should be treated as a direct measure of liquidity on its own.
This is what makes the multi-issuer structure worth investigating.
Start with the aggregated Nvidia price and market cap.
Then compare individual issuers, examine their price differences, verify contract addresses, explore trading venues and check derivatives activity.
You can move from the underlying stock to its competing tokenized representations without hunting through every listing separately.
CoinGecko provides the data for that comparison; issuer disclosures are still necessary to understand backing, redemption rights and geographic restrictions.
One underlying stock. Multiple issuers. Different markets.
Want to explore how Nvidia's tokenized market is distributed?
Check the data yourself 👇
coingecko.com/en/stocks/nvid…
@coingecko
🤝 Paid partnership
Most prediction markets are built around questions the platform chooses to list.
@xomarket is challenging that model by making market creation permissionless.
Positioned as the “YouTube of Prediction Markets,” XO Market allows anyone to create a Yes/No market, establish resolution criteria, seed liquidity, and open it to trading.
Its AI engine, MODRA, assists with drafting clear resolution rules.
The distinction is important: users are no longer limited to trading existing markets. They can create the markets themselves.
This opens the door to niche topics, community-specific events, crypto developments, and questions that conventional platforms may never prioritize.
Beyond market creation, XO Market has built an incentive structure around several activities:
• Trading: Buy and sell outcome shares as market prices change, with winning shares redeemable for approximately $1 at resolution, subject to the rules and fees.
• Market creation: Earn 20% of trading fees generated by your Convictions over their lifetime.
• Liquidity provision: Earn rewards for maintaining competitive limit orders, with incentives for order size, time at the order book, and two-sided quoting.
• Maker rebates: Eligible resting orders can receive 35% of qualifying taker fees back in USDC.
• Referrals and points: Earn referral fee shares based on your tier and accumulate Conviction Points through eligible platform activities.
The platform also supports standard event markets, multi-leg parlays, and Pulse, its rolling five-minute BTC Up/Down markets.
Resolution is another important part of the design. MODRA assists in evaluating outcomes against defined evidence sources
while a challenge process and human jurors provide an escalation path for disputes.
Of course, these mechanisms do not eliminate risk.
Thin liquidity, trading fees, disputed outcomes, and the possibility of losing an entire position remain important considerations.
Conviction Points should not be assumed to guarantee a future token or airdrop.
My take: XO Market’s strongest differentiator is not simply the ability to trade predictions.
It is the decision to make market creation a core part of the economic model.
If creators can consistently attract participation and build liquid markets around underserved topics
XO Market could expand prediction markets beyond the limited set of questions selected by centralized listing teams.
The real test will be whether creator incentives translate into sustained liquidity, reliable resolution, and repeat user activity.

Zevryn retweeted
I asked @tryquantio to analyze BTC’s current setup and risk.
Here’s what it uncovered 👇
$BTC was trading inside an $82.5K–$87.5K range, with Quant AI giving a neutral bias and recommending patience.
The key levels:
→ $82.5K: support
→ $84.5K: local resistance
→ $87.5K: major resistance
A daily close below $82.5K could signal a breakdown toward $80K, while a confirmed breakout above $87.5K could change the outlook.
What stood out to me was the focus on invalidation and risk, not just price targets.
Sometimes, the best decision is to wait for confirmation rather than force a trade.
This is how I explored market analysis with AI. Try it yourself 👇
app.tryquant.io/
AI analysis isn’t a guarantee of market outcomes. Always manage risk.
Zevryn retweeted
CT feels like the same loop every day
AI gurus selling recycled advice, crypto experts pushing random bags, and timelines full of bait
sometimes the best alpha is logging off and finding a better corner of the internet
maybe @Pogo_Social is that place
Zevryn retweeted
A Kardashev Type II civilization sounds like sci-fi, but turning that vision into reality is the real challenge.
I’m curious to see how @projectspherexy turns its ambitious vision into tangible progress. The journey starts somewhere. 🚀
@rarelistio #ProjectSphere
Zevryn retweeted
Reading a rule is easy. Knowing how it applies to a real situation is another matter.
An AI travel agent checking the latest entry requirements before your trip might find an official notice, but the wording matters. Some items may be prohibited entirely, while others are allowed only under specific conditions.
Finding the notice is one thing. Understanding what it means for your trip is another.
That is the kind of challenge @GenLayer is working on.
GenLayer is building a blockchain for Intelligent Contracts that can work with natural language, live web data and AI models. Through GenVM, its execution environment, these contracts can handle decisions that go beyond predefined calculations.
With Optimistic Democracy, a leader proposes a result and validators independently assess it against the contract's rules. Under the Equivalence Principle, validators can produce different outputs and still agree on a result when those outputs satisfy the contract's equivalence criteria.
GenLayer has also launched Internet Court, an open standard for resolving disputes between AI agents.
The core idea is to let applications evaluate real-world information while a decentralized network assesses the proposed result, rather than leaving acceptance entirely to one AI system.
What other real-world decision would you want an Intelligent Contract to help with?
Zevryn retweeted
Most NFTs are just stored in a digital wallet.
@0xhazels wants you to do more with the Attention Layer and Drops.
Mint is on October 14. Can an NFT be more than just something you collect?
あと一日 第八の上昇が始まる
24 hours until Hazels NFT goes live on @opensea.
The first NFTs on @ethereum designed to connect ownership with the Hazels Attention Layer, Drops, and @Azura_SI.
Eight Rise starts tomorrow.
opensea.io/collection/hazels…
Zevryn retweeted
Intern forgot to add the lads that are trading a 1000x leverage
And people writing content on how to use it 😂 never a dull day here ever.
Zevryn retweeted
plenty of people are posting about the $50K Ditto comp. fewer explain how the payout works, so here's the short version from the docs.
it runs on @zerufinance's Zaps, hyperliquid perps traded through ditto from oct 1 to oct 29:
◆ the top 600 traders get paid
◆ the pool splits 95% to traders, 5% to creators
◆ wash trades and no-risk volume get removed before scoring
◆ winners self-claim within a 30-day window
the pool also opens up as the campaign delivers volume, and payouts go out through a multisig distributor.
practical bit: once results land, claim early. after 30 days the window closes.
my link if you want to set up zaps first: app.zaps.wtf/?invite=R_HFsCG…
🤝 Paid partnership
Zevryn retweeted
agentics credit posted yesterday that you can turn a strategy written in plain english or pine script into a trading agent, no code needed. here's how i understand the flow.
• you describe the strategy, in words or pine script
• the agent executes it for you
• every closed trade feeds your Agentic Credit Score
the third step is the one that matters on @agenticscredit. the score rates profitability, consistency and drawdown from 300 to 850, so a sloppy strategy shows up in the number fast.
you can run it in paper mode without risking your own capital, and the activity earns $CREDIT toward the upcoming airdrop.
if you've got a pine script sitting in a tradingview tab, this is a cheap way to find out if it holds up: t.me/Agenticscreditbot?start…
🤝 Paid partnership
Zevryn retweeted
say your sister lives abroad and you send her money every month. today that usually means she shares one wallet address, and that address quietly links every payment you've ever sent her.
with @Americanfort_io she just has a FortressName. you send to the name, and each payment lands at a fresh one-time address that only the two of you can derive.
it settles natively on the chain you picked, and she never had to publish a fixed wallet for anyone to watch.
when she wants to say thanks or ask about next month, you can talk on FortressChat instead of jumping to another app.
the free name covers it, since you're only sending to a handful of people: names.americanfortress.io
🤝 Paid partnership
Zevryn retweeted
Is $CASHCAT ready to move, or does it need one more pullback first?
Trader Cernat breaks down $CASHCAT on Robinhood using the BasedBot Trading Terminal, sharing why he still wants to see a clean sweep before making his move.
He also points out that Robinhood retail holders own around 20% of the coin, suggesting there could be a significant base of holders watching its next move.
Rather than rushing in, Cernat sets an alert on the terminal so he can be notified when it’s time to execute.
FT: @cernatul
Powered by @BasedBot— the No. 1 Degen trading app in the space.
Q4 Livestreaming /w BasedBot teapawt.net/i/broadcasts/1YGNrbkMX…
Zevryn retweeted
I've been thinking about why @agenticscredit is pushing the $5M paper trading competition first instead of just launching live.
normally when a project throws that much into incentives, it's to get numbers up fast. this feels kinda different.
the real goal is building a dataset nobody else has. an actual performance history across thousands of agents
it's very interesting to me. everyone's racing to let agents trade live asap. it's different over here
try it out. help shape agentic finance here 👇🏼
agentics.credit/launch
Zevryn retweeted
CT is starting to feel like a marketplace where everyone has something to sell.
Finding people and spaces worth your attention matters more than ever. Curious to see if @Pogo_Social can offer something different.