Threading on the Edge

Amazon employees are using an internal AI tool called MeshClaw to automate unnecessary tasks to inflate their token consumption on company leaderboards.
The company set targets requiring 80% of developers to use AI weekly, then started tracking token usage on internal leaderboards earlier this year. Amazon says the data will not affect performance reviews, but multiple staff confirmed managers are monitoring it anyway. MeshClaw can initiate code deployments, triage emails, and interact with Slack. Amazon is spending roughly $200 billion on AI capex this year. Meta employees have been engaged in similar "tokenmaxxing" behavior on their own leaderboards.
https://www.ft.com/content/8ee0d3ef-9548-422d-8ff1-ebd48ad4b2ca

the most vertically integrated player always wins.

the point i want to make here is, in lieu of all these ai supercycle we are seeing, its analogous to $HYPE's success with its vertical intregration across the finance stack

also, i believe we will see the winner for the AI super cycle being the one that wins on verticalization of the ai sector stack

the ai stack consist of 3 components: apps, infra, semi

we will see history rhymes in tech supercycles

internet's biggest winner is $GOOG ~$3 trillion. near-monopoly in search. owns the crawlers, the index, the ads, the browser, the OS on half of mobile. deeply vertically integrated.

mobile's biggest winner: $APPL ~$2.5 trillion. owns the chip (M-series, A-series), the OS, the hardware, the app store, the services layer. the most vertically integrated company in history.

social's biggest winner: $META ~$2 trillion. dominant but lost roughly $1 trillion, some argue, by not owning the hardware layer. had to buy Oculus to catch up. less integrated.

cloud's winner: nobody. $AMZN, Azure, GCP share it as an oligopoly. the least vertically integrated supercycle produced the least concentrated outcome.

we have not see an AI winner yet and that AI's winner will be whoever is most vertically integrated.

that's why its the race season to own semis + infra + model + application simultaneously

Google owns TPUs, GCP, Gemini.
Microsoft owns Azure, has deep OpenAI integration.
Amazon owns custom silicon, AWS, Bedrock.
Meta owns MTIA, Llama, the distribution.

@OpenAI is the outlier, they own the model and the application but rent the infrastructure.

That's the Microsoft relationship. Whether that's a strategic advantage (focus) or a structural weakness (dependency) is the defining question of the next five years.

i believe $HYPE too will win similarly to how these hyperscalers win $GOOG $APPL $META $NDVA in trillions in market cap and its verticalization.

history says so. bet accordingly.

https://x.com/arndxt_xo/status/2054379967056851387

I think, $POD from
@dphnAI
is one of the most interesting decentralized AI runners right now: uncensored open-source models, powered by DePIN.

Launched in 2023 by Eric Hartford and team, Dolphin has already shipped dozens of uncensored models on Hugging Face with millions of monthly downloads.

Now, post-rebrand, they’ve released the full peer-to-pool tokenomics.

Here’s the no-BS, 30s breakdown.

https://x.com/Eli5defi/status/2054079896981532753

🆕 #HYPE #ETF
21Shares has launched the spot Hyperliquid ETF on Nasdaq.

Free Academy & VIP Access

Japan Is Financing the AI and Macro Supercycle (or Bubble). Everyone Thinks Is American.

Today, I want to look into 2 trades holding the global macro economy together right now.

The first is the Japan carry trade.
The second is the AI capex supercycle.

They look unrelated on the surface. One is about the yen, the BOJ, and sovereign debt sustainability. The other is about GPUs, data centers, power, and the next computing platform. But in practice, they are part of the same market structure.

Both depend on one assumption, where liquidity remains orderly

Japan has kept interest rates extremely low for years.

https://x.com/arndxt_xo/status/2054129770133356645

https://threadingontheedge.substack.com/p/japan-is-financing-the-ai-and-macro

A benchmark framework that beat all other models in real time trading @Polymarket

Best part is that it's based on an open source LLM (called MiMo)

Interesting reasons for its outperformance that the paper points to, are:

1. high confidence bets needed news support. Wrong overconfident bets were punished 😬

2. It doesn’t just read the market odds, it also processes news, order-book snapshots, and other real world signals

3. models were pushed to skip weak bets and act only when their view strongly beat market odds

4. It can analyse multiple markets at once, comparing strategies and confidence across them to avoid mistakes and maximize gains

Really interesting read 👇

https://x.com/intelscout/status/2053494253654036509?s=46&t=hr2fbvcHJGpvp_SbDstdzg

INTEL: TRUMP REMARKS

-Iran proposal unacceptable
-I have a plan on Iran
-Iran can't have nuclear weapon
-Doing service to the world

Twitter

There will be an endgame tech stack and @samoyedali research on the institutional architecture shed some light on this.

The endgame is not everyone becoming a crypto user.

The endgame is every fintech, marketplace, AI agent, and platform quietly becoming a crypto settlement company underneath.

https://x.com/arndxt_xo/status/2053850308040884388

Gate has become Polymarket's third-largest trading volume distribution channel, ranking third among Builders in both weekly and monthly trading volume.

The model of prediction market infrastructure + CEX-level distribution channels has previously been successfully validated by Robinhood + Kalshi.

This model is now being replicated on Polymarket, and we look forward to seeing more CEX-level products integrate.

https://x.com/defioasis/status/2052775518597193748

"Retail investors are still betting on gut instinct, while smart money has already started reading the market with data." 🚬

Today, we're officially open-sourcing polymarketIntel 🎯

This is a Polymarket intelligence dashboard built on my API site, and also a complete open-source example:

With the lightest code, it organizes Polymarket's markets, events, traders, order books, and real-time trades into a visual monitoring station.

https://x.com/0xInChain/status/2052735838514237605

11 tokens and promising desci projects

it will be an interesting week for desci <> icm <>
$BIO eco <> peptides

👇

https://x.com/arndxt_xo/status/2053715690016924073

INTEL: Over 600 current and former OpenAI employees reportedly sold $6.6B in stock, averaging about $11M each

Twitter

Private crypto capital is mostly moving into infra.

Here's some companies to track ↓

https://x.com/Defi_Warhol/status/2052785121124852052

🔥 Who’s About to Explode Prediction Markets on #Hyperliquid ( $HYPE)?

I think HIP-4 has had a seriously strong launch.

After less than one week since going live, the numbers are already speaking volumes and this is with just one binary “BTC above X” outcome contract per day.

https://x.com/Karamata2_2/status/2053327743643996625

the bears believe that the internet of 1999 and the ai of 2026 are highly isomorphic in market behavior.

the bulls insist that ai’s capital expenditure demands are real and verifiable, fundamentally different from the internet companies back then that burned cash in droves without viable business models.

https://x.com/arndxt_xo/status/2053445847619445054

The AI supercycle will last 15 years. We're in year 3.

Most investors are still buying Phase 1 names while the real money is already rotating into Phase 3.

I mapped the entire cycle into 4 phases with the tickers that matter at each stage:

https://x.com/cryptorand/status/2052795669426172219

All new chains use incentives. But never has a new chain managed to earn so much revenue so fast.

And these numbers are going to keep going up. On Aave, USDE is at cap, USDM nearly at cap. There is huge demand for Megaavethena.

https://x.com/lurkaroundfind/status/2050912267319808356

Price manipulation now happens almost every week, with $LAB by @LABtrade_ becoming the latest pump-and-dump token while Bitget continues playing the usual CEX role.

The LAB team, @vsadkovv, appears to control a significant portion of the supply. Wallets linked to the team still hold large allocations:

0x7Cfd8d2d8626B287bEA569b5e65AB5CBb75E9265
0x78a79D0fa0Eaf58741f5Bde7E05b5CC8F33D24d3
0x36FC85Ec486C254c9564d66de8c4210a1A20C291
0xf79ff8a5052E969a6d13E18c4E439fE5202B02Fa
0xB4b74D63F30076870d54aB9E8E6a7D18293273c3
0xe03722dedBf090Ad7A1C8F82ceB86637053E21dd

On April 8, a wallet linked to the team (0xe037) deposited 40M LAB worth $13.6M to Bitget:
0x77156a0a621d2Ac7A075C0AC3172707C2e4aa191

The LAB price started pumping on May 1, but a week earlier wallets linked to the team deposited 96M LAB worth around $63M to Bitget:
0xDd77BFbDc11Cd37fD255AE35A4ac39Df1F9d570a
0x6593aa6c31C88397c37f71259625EC92Fe4EE0bF

This looks coordinated. Gas fees (0.14 BNB) were distributed a week earlier

0x50f2760fd5E6d546EE7dcEB617F33497A3C38593
0x0559694BbB47dbA8Bc3B7ac93004EF401F2da16d

The wallet below has also been aggressively buying $LAB on-chain and depositing to Gate and Bitget, including tokens like $SkyAI, which surged 1000% in the last 30 days:

0x11fc12b988933966688d33B70651B5f2f450963C

It has been weeks since @GracyBitget promised an investigation, yet there has been no public update. If platforms cannot conduct internal investigations, identify coordinated manipulation, or provide transparency on who is behind these activities, confidence in market integrity keeps declining.

ZachXBT still has a reward open for credible intel that lead to identifying the actors behind this operation.

Stay smart.

https://x.com/specteranalyst/status/2052334885776368019?s=46&t=hr2fbvcHJGpvp_SbDstdzg