andrey
201 posts


@NaturalEye78321 Nice. Yet Thailand and Vietnam also have beautiful beaches, and the chance of getting robbed or killed are almost zero, what you cannot say about Brazil.
English

@MikeMagic80 @ExtremeFootbal4 The third one is not known by 90% of non Americans as nobody but them and the Japanese play it or know anything about its history. It's like criket or polo - a local sport.
English

@ExtremeFootbal4 Those top 3 are undisputedly goated no matter what country you come from.
Everything else is a battle for number 4.
English

@MoonwellDeFi @LukeYoungblood So, you guys decide to pause the OP market at highly volatile times when liquidations are highly probable without any announcements?
English

Moonwell on Ethereum Mainnet?
On the latest Governance Call, @LukeYoungblood shares the latest on a potential activation of the protocol on Ethereum Mainnet and asks for community feedback on the initial market selection.
English

@FefeDemeny Well, yesterday a guy almost ran over me and my wife with his car after eye-sight confirmation with her, one new problem
English

@Hellinikon @TansuYegen You have just won the "stupidest comment" award.
English

@TansuYegen Who in their right mind would go to Dagestan for vacation? The sheep are more civilized than the people 🤣
English

@CryptoSteveO1 Yeah, and one day Polymarket moves to their own blockchain and most of tge revenue will be gone, Pol will be just one of many L2s.
English

Polygon $POL is SO undervalued; its crazy!
Polygon's marketcap is $998M.
ICP's marketcap is $1.1 BILLION (VERY overvalued)
Near's marketcap is $1.67 BILLION (overvalued)
Polkadot's marketcap is $2 BILLION (VERY overvalued)
Sui's marketcap is $3.6 BILLION (VERY overvalued)
Hedera's marketcap is $3.79 BILLION (extremely overvalued)
Avalanche's marketcap is $3.9 BILLION
Cardano's marketcap is $8.9 BILLION (RIDICULOUSLY overvalued) ADA doesn't even make it in the top 20 in ANY revenue or user metric!!
Yet Polygon's user and REVENUE metrics SMOKES all of these chains. Polygon's TVL has increased the most over the last 7 days. Visa, Walmart and many other companies are onboarding the Polygon chain. Eventually the market will figure this out and price action will correct.
Source: defillama.com/chains
$BTC $ETH $BNB $SOL $ICP $NEAR $DOT $HBAR $AVAX $SUI $ADA

English

@leonid_krivenko @SpiritBroDude @vivoplt A very narrow minded thing to say. Of course people use local AI in Russia, Alisa is the default assistant for years now, and Sber's Kandinsky is a solid model for text2image being more esteticaly pleasing, then Stable Diffusion 1.5
English

@SpiritBroDude @vivoplt Nobody uses them in Russia. Usually they either use chatgpt, or if they don't have money to pay for unlimited version, they use deepseek after reaching the limit on chatgpt
English

@R1chardMaur1ce @NoLimitGains I would not be surprised if in this context "partnership" means buying a few GPUs so they can run LLMs locally in their office.
English

Qwen vs Qwen: Who Overthinks the Most?
Last round, we found out Qwen was the overthinking champion across all our favourite open-source models. So this time... I pitted Qwen against itself. Who burns the most tokens?
🥇 The overthinking king: Qwen3.5 9B
> This little guy went absolutely feral on tokens:
> 17,386 tokens on SEND+MORE=MONEY ❌ still got it wrong
> 13,941 on a prime triplet question ❌ wrong again
> 8,231 on n²+n+41 ❌ yep, wrong too
> Topped 4 out of 5 questions in token spend
> The only model to fail, and it failed 3 out of 5. Thinking long ≠ thinking right.
🥈 Runner-up: Qwen3.6 35B A3B
> The local crowd favorite. Always top 2 in token spend, but unlike the 9B, it nails every single answer.
🟢 The chill ones
> Qwen3.6 Plus and both 27B dense models (3.5 & 3.6) keep it short and confident. Even Qwen3.5 397B A17B, a massive MoE, barely thinks. So no, scale doesn't mean overthinking.
🔍 3.5 → 3.6: the glow-up is weird
> 35B A3B: 3.6 thinks ~2x longer than 3.5 (9,798 vs 4,966 on Q1). Newer = more reasoning.
> 27B dense: basically tied, 3.6 slightly leaner. Feels like the 3.6 MoE was trained to think harder. The 3.6 dense was trained to think smarter.
📊 The pattern:
> Small models and small-active MoEs (9B, 35B A3B) overthink the hardest.
> Big dense and big-active MoEs are way more efficient.
ℹ️ Running Qwen locally? Heads up:
> 35B A3B → fast tokens/sec, but pays in reasoning length
> 27B dense → slower per token, but finishes first because it just thinks less
You save compute on params, you spend it on thinking. There's no free lunch.
English

@TedPillows Interesting to see if he can convert it to cash. In my view of the world - his account should be blacklisted accross the CEX'es and any other bridges to fiat
English

@JustDeauIt There is also ASTER, the second largest perp dex which, when I last checked, is not far behind in terms of metrics, yet the token is near ATL, probably because the buybacks are not like on HYPE. Please consider comparing the two and/or including it in the watch list.
English

Not sure I've ever seen a crypto asset perform the way HYPE has in a bear market.
Fundamentals are in decline, relative to valuation.
Said another way, *the market is paying up for each dollar of sales in the bear market.*
That's an anomaly, especially in the first half of bear markets, when valuations typically drop much faster than fundamentals (using 365-day lookbacks, when market conditions were stronger).
This can be either bullish or bearish.
I'm sharing a full quarterly ecosystem update tomorrow with readers of @the_defi_report, including:
- Perps DEX
- HIP-3
- HyperEVM
- Token Economics
- Valuation
- Risks
If you'd like to have the (free) report + dashboard dropped into your inbox when it's published, see the link below 👇


English

@haydenzadams Today: 5k eth/usdc position on 30% range bringing in 5 usd daily.
English

@EricTrump @worldlibertyfi So, creating a rug pull on a memecoin us something not ridiculous and to be proud for?
English

The only thing more ridiculous than this lawsuit is spending $6 million on a banana duct-taped to a wall. We are incredibly proud of the @worldlibertyfi team…
Zach Witkoff@ZachWitkoff
Justin Sun’s recent lawsuit against @worldlibertyfi is a desperate attempt to deflect attention from Sun’s own misconduct. His claims are entirely meritless, and World Liberty looks forward to getting the case thrown out promptly. He engaged in misconduct that required World Liberty to take action to protect itself and its users. World Liberty will continue to take all necessary steps to protect its community.
English

Qwen3.6 27B is now in LM Studio!
Vision, reasoning, and agentic tool calling - running locally on your computer.
Surpasses previous Qwen models many times its size
🚀👾🔥
lmstudio.ai/models/qwen/qw…
English

@TonySeverinoCMT Looks like the mm's are paying the idiotic kols like martyparty to pump and dump
English


















