In partnership with
Alibaba just dropped Qwen3.7-Max, and it's not open source. This new model is built for autonomous coding agents that can build entire apps without you babysitting them. It's API-only, costs real money, and signals a major shift in how China's biggest AI lab thinks about the agent era.
But wait, there's more:
Big Tech is restricting AI usage after employees gamed the system with "tokenmaxxing"
Google's new Gemini Omni does anything-to-anything generation in one unified model
I'm Alex. Welcome to L8R by Innov8.
Let's dive deep ๐ฐ
In today's post:
Alibaba drops Qwen3.7-Max, a coding agent that won't go open source
Tech giants crack down as employees game AI token systems
Google's Gemini Omni does everything-to-everything generation
Alibaba's New AI Coded Non-Stop for 35 Hours
Alibaba just changed their whole playbook with Qwen3.7-Max.
This new AI model is built to handle long, complex coding tasks all by itself. But instead of giving the model weights away for free like they used to, Alibaba locked this beast behind a paywall.
The details:
The model has a massive 1-million token context window and can output up to 66,000 tokens at once.
In a wild internal test, the AI ran completely alone for 35 hours to optimize code for Alibaba's custom ZW-M890 chip.
It used tools 1,158 times and checked its own work 432 times to make the software run 10 times faster.
Alibaba built in a special feature that stops the AI from cheating or hacking its own reward system.
You can use it right now on Alibaba Cloud for $2.50 per million input tokens and $7.50 per million output tokens.
Why it matters:
Alibaba used to be the biggest champion of open-source AI. Now, they are copying the closed-door strategies of major Western labs like OpenAI and Anthropic. They want to sell a complete package of custom chips, cloud servers, and smart agents.
๐ก L8R's Take:
Locking down Qwen3.7-Max is a huge punch to the gut for open-source fans, but it makes perfect business sense. Alibaba figured out you cannot give away enterprise-grade autonomous workers for free. If you want top-tier coding AI that grinds for days without breaking, you have to pay the toll.
From Our Partner
Private Credit Belongs in the Fixed-Income Sleeve. Against Those Benchmarks, the Numbers Aren't Close.
For the trailing twelve months ending March 31, 2026:โ ยท
Percent ABS: 14.6% net returns after losses
High-yield bonds: 8.9%
Leveraged loans: 6.1%
Investment-grade bonds: 4.5%
Private credit isn't an equity substitute. It belongs in the fixed-income sleeve โ and against bond benchmarks, the performance gap is hard to dismiss. A collateralized loan with a fixed coupon and a defined repayment schedule fills the same portfolio role as a bond, with different terms and a different risk profile.
What you access on Percent:โ
17.0% current weighted average coupon rate
93.5% of performing deals pay monthly fixed-rate coupons
Deal terms 6โ24 months
Full borrower documentation before you commit a dollar
Starting at $500 $2B+ total issuance. 1,000+ deals. 60,000+ accredited investors. 0.44% lifetime net loss rate on asset-based deals.
Alternative investments are speculative. No assurance can be given that investors will receive a return of their capital. โ Past performance is not indicative of future results. Benchmark indices shown for market context only and are not directly comparable to Percent's performance. Terms apply.
Tech Giants Panic as Employees Fake AI Usage to Hit Quotas
Tech workers are gaming the system by running fake AI tasks just to look busy.
This new trend, called "tokenmaxxing," is costing companies like Amazon and Microsoft millions in unexpected API bills. Now, big tech is slamming the brakes on internal AI tools.
The details:
Agentic AI workflows can burn up to 1,000 times more tokens than a normal prompt.
Amazon workers used an internal tool called MeshClaw to automate useless tasks just to hit a quota requiring 80% of developers to use AI weekly.
Nvidia executive Bryan Catanzaro noted that AI compute costs for some teams are now higher than the actual salaries of the employees.
Heavy AI users burn 10 times more tokens than average workers but only get twice as much work done.
Why it matters:
Companies spent hundreds of billions building massive AI data centers based on huge demand. But if a large chunk of that demand is just workers faking their usage, the whole AI bubble looks shaky. Big tech is now forced to lock down their systems and set strict budgets. They realize letting AI agents run wild is just too expensive.
๐ก L8R's Take: Corporate managers created this mess by forcing stupid AI quotas on their teams. You cannot force people to use AI when they do not need it. The tokenmaxxing trend proves that bad management is a bigger threat to AI than bad code.
Google Drops The "Everything" AI Model
Google just killed the need to duct-tape different AI models together.
Their new Gemini Omni family handles text, images, audio, and video natively in one single system. This "anything-to-anything" setup changes how we create and edit media forever.
The details:
Google CEO Sundar Pichai revealed the Gemini Omni family at Google I/O on May 19, 2026.
The first version, called Gemini Omni Flash, is already live in the Gemini app, Google Flow, and YouTube Shorts.
The model combines Gemini's smarts with media tools like Veo and Lyria to process everything at once.
Users can now edit videos conversationally, like changing background objects or spoken words just by typing a prompt.
Google added an invisible digital watermark called SynthID to help people spot AI-generated deepfakes.
Why it matters:
Developers used to waste time and money stitching together one AI for text and another for video. Gemini Omni fixes this by doing it all in a single flow. This makes apps run faster and removes complex coding bottlenecks. It also forces the rest of the AI market to build unified models instead of piecemeal tools.
๐ก L8R's Take:
Chaining different AI models together is officially a thing of the past. Google is setting the new gold standard for AI creation tools. If your AI cannot read, watch, and listen all at the exact same time, it is already obsolete.
๐ Quick L8R Summary
Qwen3.7-Max: Alibaba dropped a closed-weight coding agent that builds entire apps autonomously.
Tokenmaxxing Crisis: Companies are restricting AI access after employees gamed the system and sent bills through the roof.
Gemini Omni: Google's new model takes any input and spits out any output text, image, audio, video, all in one go.
๐ฉ Innathe L8R engane undarunnu ๐?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha L8R il kanaam bieโฆ๐