The Transcript
RT @TheTranscript_: $UBER CFO @_balaji_km: Uber's profit engine lies in smaller U.S. markets, not top cities

"this is a very, very common misconception. We've heard many times that Uber's profit pools are concentrated in the top cities, and it could not be further from the truth." https://t.co/KvEaIn4beM
tweet

Moon Dev
Why I Fired My Dev Team: Using Claude’s New Agent Squads to Out-Code a $200k Salary

coding as we know it just ended because we finally moved from single agents to teams that can actually argue with each other to find the truth. most people are still stuck prompting a single window and waiting for a response but the real power shift is happening in the background where anthropic just dropped a way for machines to orchestrate themselves. i used to spend hundreds of thousands on developers for my apps because i thought i was not capable of doing it myself but now i realize that code is the ultimate equalizer if you just know how to point the tools in the right direction

you think you are fast now but wait until you see what happens when the ai stops waiting for your permission to move to the next task. the secret lies in a new experimental feature called agent teams which is essentially a squad of independent claude instances that share a task list and talk to each other directly. unlike the standard sub agents that just report back to a main boss these teammates have their own context windows and can challenge each other findings which means they catch bugs before you even see the first line of output

everyone wants to know how to actually use this in a way that makes money instead of just playing with new tech in a vacuum. if you have been following my journey you know i lost plenty of money to liquidations and over trading before i decided to automate everything because humans are just too emotional for the markets. this is why i decided to build a massive management system for my claude code sessions right inside my quant app because i was getting tired of losing my mind in twenty different terminal tabs. the problem with the current way we use these tools is the lack of organization which is the silent killer of any serious trading system

automation is the only way to survive the slaughterhouse of the modern markets but the tooling has to match the ambition. i am building this manager using electron and xterm js to create a real terminal experience that allows me to run multiple claude sessions in a flexible grid layout. instead of clicking through a mess of windows i can now see exactly what is happening across my entire stack in a four by four grid that shows me which agent is currently cooking and which one has finished its task. i even had the ai build a feature where the terminal border glows a soft yellow while it is working and then flashes green when the output is done so i can keep my eyes on the data

the window of opportunity we are in right now is completely subsidized by the big labs and if you are not taking advantage of it you are essentially leaving the greatest leverage in history on the table. people complain about two hundred dollars a month for a pro plan but they do not realize that we are basically getting free intelligence compared to what it will cost once the market settles. i am running these agent teams to build features in eight minutes that used to take me a week of back and forth with expensive contractors. it is all about the iteration to success mindset because you have to be willing to kill your old code the second a better path reveals itself

if you are still manual trading or manual coding you are basically bringing a knife to a drone fight and wondering why you keep getting liquidated. my name is moon dev and i believe that anyone can reach the top if they are willing to learn live and automate the boring parts of their life. we are keeping the interface minimal and developer focused with no bloat and no distractions because the goal is to earn by learning faster than the competition. the system persists all my project directories so i never have to setup my environment twice which saves me hours of cognitive load every single week

most people will see this new team feature and just keep using it for basic chat completions instead of building a whole management layer on top of it. we integrated hyperli[...]

quid data layers and liquidation trackers right next to the code terminals so i can monitor the market while the agents are building my next strategy. this is the convergence of high frequency data and high frequency development where the bottleneck is no longer the syntax but how fast you can think of the next experiment. learn baby learn so you can earn baby earn is the only philosophy that matters when the tools are moving this fast

the agents are now able to handle front end tasks and terminal setup in parallel without stepping on each other toes which is a massive leap forward. i had one agent working on the preload scripts while another was styling the collapsible sidebar and they coordinated the integration testing without me having to intervene once. this is why i say the tooling is now more important than the model itself because even a smart model is useless if it cannot interact with the real world efficiently. we are building a downloadable system that gives us institutional grade power from a home setup and that is the real magic of this era

don't get distracted by the hype of which model is one percent smarter when the real gains are in how you manage the agents you already have. i am sticking with anthropic for now because their tooling is light years ahead of the competitors who are just focusing on chat interfaces. by building a dedicated space for my code inside the quant app i have removed the friction that leads to procrastination and mistakes. it is a serious game and the markets do not care about your feelings so you might as well build a system that can outwork any human on the planet

the true equalizer is not just having the code but having the systems to manage that code at scale without losing your mind. we are just getting started with these agent teams and i can already see a future where i am managing an entire company worth of output from a single dashboard. stay focused on the fundamentals and keep moving forward because if you cannot fly you should run and if you cannot run you should crawl. by all means keep moving because the automated future is already here for those who are willing to build it
tweet

Illiquid
Following Nikkei's many articls, WSJ is now profiling Nittobo too. Resonac and Ajinomoto also get shout-outs.

https://t.co/yOvn0QvJQB
tweet

God of Prompt
RT @rryssf_: MIT researchers just mass-published evidence that the next paradigm after reasoning models isn't bigger context windows ☠️

Recursive Language Models (RLMs) let the model write code to examine, decompose, and recursively call itself over its own input.

the results are genuinely wild. here's the full breakdown:
tweet

Lumida Wealth Management
1/ Daily News Round-Up:

- Markets rebound after AI-driven tech selloff
- Chinese factory in Ohio tests onshoring limits
- AI accelerates in healthcare across diagnostics & ops
- China signals banks to cut US Treasury exposure
- Big Tech to spend ~$670B on AI in 2026
tweet

Moon Dev
🚨 Openclaw for Tradingview https://t.co/5xSdqi0ng5
tweet

Moon Dev
🚨 Openclaw for Tradingview https://t.co/CZKhckNpCo
tweet

Bourbon Capital
RT @BourbonCap: Stocks to Buy and Hold for the Next Decade

1. $AMZN - Amazon

Amazon’s global fulfillment and last-mile logistics network is nearly impossible to replicate. Years of infrastructure investment have created a moat that supports fast delivery, low unit costs, and unmatched scale.

AWS remains the backbone of global cloud infrastructure, now operating at a $142 billion annual revenue run rate. It offers the broadest functionality, security, and ecosystem in the market, powering the majority of enterprise and government cloud transitions. More than 90% of the top 1,000 customers use Graviton, which delivers up to 40% better price-performance.

Generative AI has already become a multi-billion-dollar growth engine for AWS. Bedrock reached multi-billion-dollar ARR with 60% QoQ spend growth, while Trainium2 is ramping faster than any internal chip in AWS history, offering 30–40% better price-performance than GPUs.

Amazon’s 250+ million Prime members globally create recurring revenue, high retention, and powerful cross-selling leverage across retail, media, and services.

The company’s grocery and quick-commerce expansion continue to scale. Grocery now spans 1,000+ U.S. cities, with a target of 2,300 locations, while Amazon Now tripled Prime shopping frequency in India and is showing strong early results in the U.S. and U.K.

Advertising is another major growth driver. Prime Video ads now reach 315 million viewers across 16 countries, up from 200 million previously. AI tools such as Ads Agent and Creative Agent are improving advertiser ROI and accelerating campaign creation.

Thanks to its unmatched ecosystem, Amazon is widely expected to become the first company to surpass $1 trillion in annual revenue by 2028, underscoring its ability to monetize across multiple industries.
tweet

Benjamin Hernandez😎
Good morning.

The pre-market session is active with $GLW, $TTWO, and $ORCL showing strength. We are also monitoring $SQ and $SHOP for breakouts.

Make sure you are not over-trading the morning volatility. Stick to your proven strategy. ✅ https://t.co/gxnsQrzUuM
tweet

Startup Archive
Elon Musk: “Failure is irrelevant unless it’s catastrophic”

Elon is asked why he’s so tolerant of the failure that often stems from taking big risks, to which he responds:

“I think of these things as: There’s a certain amount of time, and within that time, you want the best net outcome. For the set of actions you can do, there will be some which will fail and some which will succeed. And you want the net useful output of your actions to be the highest.”

He uses baseball as an analogy:

“You can’t just sit there and wait for the perfect pitch . . . So what you’re really looking for is: What’s the batting average? There’s going to be some amount of failure. You just want your net useful output to be maximized. Failure is essentially irrelevant unless it’s catastrophic.”
tweet

God of Prompt
RT @godofprompt: Remotion-style motion videos used to require coding, CLI setup, and hours of debugging.

Now you just type a prompt.

@TopviewAIhq just launched Vibe Editing in beta and it's exactly what it sounds like:

> Describe the video you want
> Upload your assets (or don't)
> The AI agent builds the whole thing in your browser

No local installation. No environment setup. No timeline editing.

Product launches, explainer videos, promo clips... all from a text prompt.

This is what "vibe coding" looks like for video.

Now launched on beta.

Try it!
tweet