Rishi

Engineer | Learning ~ GenAI • RAG • Agents

[Rishi @ByteBackwards]

Everyone is talking about SpaceX's $1.75T IPO valuation.
But I think the most interesting part isn't the rockets.

It's how modern market structure, passive investing, and index inclusion could influence one of the largest IPOs in history.

I dug into the numbers and wrote about why the SpaceX IPO may be as much a Wall Street story as a space story.

Read here: https://t.co/VDLDOHI89A

[Olivia Chowdhury @Oliviacoder1]

MIT just made every AI company's billion dollar bet look embarrassing.

They solved AI memory. Not by building a bigger brain. By teaching it how to read.

The paper dropped on December 31, 2025. Three MIT CSAIL researchers. One idea so obvious it hurts. And a result that makes five years of context window arms racing look like the wrong war entirely.

Here is the problem nobody solved.

Every AI model on the planet has a hard ceiling. A context window. The maximum amount of text it can hold in working memory at once. Cross that line and something ugly happens — something researchers have a clinical name for.

Context rot.

The more you pack into an AI's context, the worse it performs on everything already inside it. Facts blur. Information buried in the middle vanishes. The model does not become more capable as you feed it more. It becomes more confused. You give it your entire codebase and it forgets what it read three files ago. You hand it a 500-page legal document and it loses the clause from page 12 by the time it reaches page 400.

So the industry built a workaround. RAG. Retrieval Augmented Generation. Chop the document into chunks. Store them in a database. Retrieve the relevant ones when needed.

It was always a compromise dressed up as a solution.

The retriever guesses which chunks matter before the AI has read anything. If it guesses wrong — and it does, constantly — the AI never sees the information it needed. The act of chunking destroys every relationship between distant paragraphs. The full picture gets shredded into fragments that the AI then tries to reassemble blindfolded.

Two bad options. One broken industry. Three MIT researchers and a deadline of December 31st.

Here is what they built.

Stop putting the document in the AI's memory at all.

That is the entire idea. That is the breakthrough. Store the document as a Python variable outside the AI's context window entirely. Tell the AI the variable exists and how big it is. Then get out of the way.

When you ask a question, the AI does not try to remember anything. It behaves like a human expert dropped into a library with a computer. It writes code. It searches the document with regular expressions. It slices to the exact section it needs. It scans the structure. It navigates. It finds precisely what is relevant and pulls only that into its active window.

Then it does something that makes this recursive.

When the AI finds relevant material, it spawns smaller sub-AI instances to read and analyze those sections in parallel. Each one focused. Each one fast. Each one reporting back. The root AI synthesizes everything and produces an answer.

No summarization. No deletion. No information loss. No decay. Every byte of the original document remains intact, accessible, and queryable for as long as you need it.

Now here are the numbers.

Standard frontier models on the hardest long-context reasoning benchmarks: scores near zero. Complete collapse. GPT-5 on a benchmark requiring it to track complex code history beyond 75,000 tokens — could not solve even 10% of problems.

RLMs on the same benchmarks: solved them. Dramatically. Double-digit percentage gains over every alternative approach. Successfully handling inputs up to 10 million tokens — 100 times beyond a model's native context window.

Cost per query: comparable to or cheaper than standard massive context calls.

Read that again. One hundred times the context. Better answers. Same price.

The timeline of the arms race makes this sting harder. GPT-3 in 2020: 4,000 tokens. GPT-4: 32,000. Claude 3: 200,000. Gemini: 1 million. Gemini 2: 2 million. Every generation, every company, billions of dollars spent, all betting on the same assumption.

More context equals better performance.

MIT just proved that assumption was wrong the entire time.

Not slightly wrong. Fundamentally wrong. The entire premise of the last five years of context window research — that the solution to AI memory was a bigger window — was the wrong answer to the wrong question.

The right question was never how much can you force an AI to hold in its head.

It was whether you could teach an AI to know where to look.

A human expert handed a 10,000-page archive does not read all 10,000 pages before answering your question. They navigate. They search. They find the relevant section, read it deeply, and synthesize the answer.

RLMs are the first AI architecture that works the same way.

The code is open source. On GitHub right now. Free. No license fees. No API costs. Drop it in as a replacement for your existing LLM API calls and your application does not even notice the difference — except that it suddenly works on inputs it used to fail on entirely.

Prime Intellect — one of the leading AI research labs in the space — has already called RLMs a major research focus and described what comes next: teaching models to manage their own context through reinforcement learning, enabling agents to solve tasks spanning not hours, but weeks and months.

The context window wars are over.

MIT won them by walking away from the battlefield.

Source: Zhang, Kraska, Khattab · MIT CSAIL · arXiv:2512.24601
Paper: https://t.co/bQEHMCRSEj
GitHub: https://t.co/vR6Pz8FQHD

[Rishi @ByteBackwards]

@NalinisKitchen If you redistributed all wealth equally today, within a few years much of it would return to the same people.

[Rishi @ByteBackwards]

Be delusional!

[Rishi @ByteBackwards]

@nirajxdev Revenge is a dish best served cold.

[Rishi @ByteBackwards]

@Brennan_Lup Children still learn basic math even though we have advanced calculators. Coding is becoming the same thing. AI can write code, but understanding logic, systems, and how to fix what AI breaks is still valuable.

[Rishi @ByteBackwards]

Be delusional!

[Kalshi @Kalshi]

JUST IN: Elon Musk is projected to be worth $1.39 trillion this year

[Iric Midel @IricMidel]

AI really fooled everyone to be founders when 99% end up as token consumers.

[Rishi @ByteBackwards]

@DesiDebugger @nicotwtss Best reply! mic drop💯🙌

[Rishi @ByteBackwards]

@Literariium And the longest story hides inside it.

[Rishi @ByteBackwards]

@Im_IrushiK English, and you?

[Rishi @ByteBackwards]

It's not FAANG anymore, it's MANGOS.

[Rishi @ByteBackwards]

FIFA World Cup!

[Rishi @ByteBackwards]

@WorldCupMotion It's : Spain, Argentina or England

[Rishi @ByteBackwards]

@anurag_gharat @WorldCupMotion Spain🙌

[Rishi @ByteBackwards]

I was trying to book an Uber Auto during peak Bangalore traffic. The app showed ₹180, but no driver was accepting the ride.

A local auto driver pulled up and asked where I wanted to go. I told him the location, and he immediately demanded, "₹300 fix, saar. Heavy traffic."

Instead of arguing or showing him the Uber price, I changed the story. I looked exhausted and said: "Bhaiya, I have a massive job interview in 15 minutes. If I'm late, my entire month's preparation goes to waste. I only have ₹200 cash on me right now. Please help me out."

He paused, looked at my face, waved his hand, and said, "Baitho, bhaiya. Chalo."

The funny part? It wasn't a job interview. I was just going to meet a friend. And I had plenty of money in my bank.

The Lesson: Logic and data (like showing him the Uber screen) breed resistance and negotiation. Empathy and high-stakes narratives breed cooperation. If you want someone to agree to your terms, stop fighting them with facts, give them a compelling reason to become the hero of your story.

[Rishi @ByteBackwards]

Small minds compete for position, big minds collaborate for scale.

[Rishi @ByteBackwards]

@nalinrajput23 Team. More brains usually lead to better decisions because strength is in numbers. It also helps prevent you from moving in a single direction or getting stuck in your own blind spots.

[Rishi @ByteBackwards]

Yesterday, I went to an art gallery and found myself looking at a painting that was literally just a giant blue square. A guy next to me sighed and said, "Modern art is so lazy."

I looked deeply at it, lowered my voice, and whispered:
"It’s actually a protest. The artist used a specific toxic chemical dye that is slowly fading over decades. By the time we die, the canvas will be completely blank. It’s a commentary on the fleeting nature of human memory."

He stared at it for three full minutes, visibly moved, and said, "Wow. That is profound."

I made it up on the spot. It was literally just a blue square. This reminded me of a massive psychological truth: Reason doesn't move people. Emotion and mystery do.

If you wrap a blank canvas in a brilliant narrative, people will gladly find their own meaning in it. We don't buy products or facts, we buy how a story makes us feel.

[Rishi @ByteBackwards]

@santhil_X Here to connect.🤝

[Rishi @ByteBackwards]

@ChenyuW64562111 @MIT_CSAIL @GoogleDeepMind Congratulations!🎉