In partnership with
n
OpenAI just dropped proof that AI can solve math problems humans couldn't crack for over a decade.
Their new Astra model tackled ten problems that stumped mathematicians and computer scientists since the mid-2010s.
And they backed it up with 249 pages of machine-verified proof.
But wait, there's more:
DeepSeek's new V4-Flash model brings production-ready performance for developers
Snapchat and LinkedIn are fighting back against AI-generated garbage flooding their platforms
I'm Alex. Welcome to L8R by Innov8.
Let's dive deep 🐰
In today's post:
• OpenAI's Astra cracks 10 unsolved math problems
• DeepSeek drops V4-Flash for production use
• Snapchat and LinkedIn ban AI slop
$2,000 AI Just Solved 10 Unsolved Math Problems
OpenAI just dropped a massive bomb on the math world.
Their unreleased Astra model solved ten complex math problems that humans could not crack for decades.
It proves AI is now a real research partner, not just a fancy chatbot.
The details:
OpenAI used an internal version of Astra, their next big model family built to handle complex tasks that take days.
The AI solved ten major open math problems, including a group theory question that stumped humans since 1999.
The whole process only cost about $2,000 in compute power to generate the winning solutions.
Every single proof includes a machine-checkable Lean 4 certificate to prove the AI did not hallucinate the answers.
OpenAI gave full credit to the AI for the math, saying claiming human authorship would be a lie.
Why it matters:
This changes everything for science and research.
AI is stepping up from writing simple code to making actual frontier discoveries.
The machine-checked proofs mean we do not have to worry about AI making things up.
It turns solving world-class math problems into a cheap, easy process.
💡 L8R's Take:
Astra is going to break the world in the best way possible.
If $2,000 gets you ten solved math mysteries, imagine what a million dollars will do.
Human researchers better learn to work with these agents right now, or they will be completely left behind.
From Our Partner
10x the context. Half the time.
Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.
DeepSeek Drops V4-Flash: A Dirt Cheap Agent Killer
DeepSeek just unleashed a massive upgrade to its V4-Flash model that targets serious builders.
The new 0731 version brings frontier-level smarts to autonomous agents and tool use without the insane price tag.
It gives you top-tier performance at a fraction of the cost of closed models.
The details:
The model packs 284 billion total parameters but only uses 13 billion active parameters per token to keep operations fast and cheap.
Performance jumped massively compared to the preview version, showing a 645% improvement on the DeepSWE coding benchmark.
Builders get a massive 1-million token context window and configurable reasoning levels for complex tasks.
API pricing is incredibly low at just $0.14 per million input tokens, which is about one-fifth the cost of similar proprietary models.
DeepSeek released the model under an MIT license on July 31, 2026, meaning you can download the open weights directly from Hugging Face.
Why it matters:
This release totally changes the math for AI developers.
You no longer have to pay huge fees to closed labs to get reliable agentic behaviors and structured tool calling.
By making the model open-source and highly efficient, DeepSeek lets you run top-tier AI locally or on much cheaper cloud servers.
💡 L8R's Take:
DeepSeek is completely outplaying the big tech giants right now.
They are giving away world-class models for practically nothing while others guard their secrets.
If you are building AI agents in 2026 and not testing this model, you are just throwing your money away.
From Our Partners
Auto-Generate Free SEO Audit Report
Most e-commerce sellers have no idea why their listings aren't ranking. Wrong keywords, missing metadata, weak product descriptions — the problems are there, but no one's told you where to look.
StoreClaw runs a full SEO audit across your Amazon and Shopify stores automatically. In minutes, you get a clear score, a breakdown of what's hurting your rankings, and exactly what to fix.
No manual review. No SEO agency. No guesswork.
Connect your store and StoreClaw surfaces every issue that's costing you search visibility — then tells you how to fix it.
Free to start. No credit card required.
Snapchat and LinkedIn Declare War on "AI Slop"
Nobody likes a feed full of fake, robot-written garbage.
Snapchat and LinkedIn are finally taking out the trash by blocking low-quality AI content.
They want to make sure real humans get the spotlight again.
The details:
Snapchat updated its Spotlight rules on July 31 to completely block fully AI-generated videos from getting recommended to new viewers.
LinkedIn added a brand new "Seems like AI slop" report button right on the three-dot menu of every post.
LinkedIn is also killing off its "Enhance your post" AI writing tool and swapping it for a simple spell-checker.
A recent study found that a massive 41% of long-form posts on LinkedIn were written by machines.
Why it matters:
Platforms are realizing that AI spam ruins the user experience.
The era of using cheap AI tools to pump out hundreds of daily posts for easy growth is ending.
Social media companies are drawing a hard line between using AI to help you edit and using AI to do all the work for you.
💡 L8R's Take:
It is about time these platforms woke up.
AI should be a tool to make your ideas better, not a factory that prints out boring junk.
If you rely on bots to write your posts, your free ride is officially over.
🚀 Quick L8R Summary
OpenAI Astra: OpenAI's new Astra model just cracked ten math problems that stumped researchers for over a decade.
DeepSeek V4 Flash: DeepSeek dropped their production V4-Flash model with beefed-up coding and agent skills.
AI Slop Crackdown: Snapchat and LinkedIn are blocking low-quality AI spam from flooding your feeds.
📩 Innathe L8R engane undarunnu 👇?
We read every reply - just reply to this email and let us know how we can improve !
Appo adutha L8R il kanaam bie…👋