Eric Magalhães retweeted
Managers fix behavior. The fear stays.
Every "difficult" employee is a behavior the org noticed.
Which side of the mirror have you been managing?
Book: Know Your People — knowyourpeoplebook.com
#management #leadership #knowyourpeople
Eric Magalhães retweeted
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: arxiv.org/abs/2512.24601
GitHub: github.com/alexzhang13/rlm
Eric Magalhães retweeted
Kill Overkill
Business is often seen as the art of acquisition. Acquiring talent, customers, revenue, profits, mindshare, marketshare. Building and growing requires consumption, addition, parlaying some of this into a lot of that.
But the smartest businesses — the ones that tend to stick around for the long haul — know that existence is also about avoidance.
Avoiding careening variable costs, avoiding getting involved with things that aren't core to your business, avoiding spending time on things that don't matter, avoiding bad investments, avoiding people who don't help you prosper, and even avoiding customers who aren't the right fit.
However, there's something even more fundamental to avoid. It would be easy to call it "complexity" but that's not quite it. Complexity can be necessary, and intricacy can be quite beautiful — just stare into a Moorish mosaic and you'll know it.
So complexity isn't the issue.
Overkill is the issue.
Avoiding overkill is the real cheat code. That's how you jump levels. It's how you make a lot more progress with a lot less effort.
Overkill is the dust that settles on the stuff that took a lot of energy to build or buy, but turned out not to be necessary. The over-engineered, over-designed, over-hired, over-litigated, the over-spent, over-promised, over-deliberated.
Overkill is the policy that was written but never enacted. The technology that was purchased that was never used. The seven steps that could be handled in two. The nine people in a meeting made for three. The business equivalent of the 12 bedroom house for a family of four. The cooks when you don't even have a kitchen.
Overkill is using five different products to run a single project. Overkill is an seven-stage interview process that exhausts everyone involved. Overkill is acting like a company 100x your size. Overkill is buying what they bought but that you don't need. Overkill is paying thousands for something worth hundreds. Overkill is hoping that losing more will turn into a win.
In our 24 years, there's nothing we've tried to avoid more at 37signals than overkill. In the things we do, in the way we work, in the things we buy, in the things we use. And, especially, in the products we make for our ourselves and our customers. Basecamp and HEY are built to do what they need to do, in the most straightforward, elegant, and enjoyable way, and nothing more. So you can avoid overkill too.
Every day is an opportunity to find just right. To toss that policy that's in the way. To slim down the stack. To sharpen things up. To eliminate the work that doesn't need to be done. To polish the scratched glass so you can see through again.
Amass what you need, but ignore even more. Kill overkill.
Eric Magalhães retweeted
Today we're introducing the official Rails Job Board, a platform for Rails developers to find their next role (and companies to find their next Rails developer). Get started and find your next career match here: rubyonrails.org/2023/11/27/r…
Eric Magalhães retweeted
Hello, Brazil 🇧🇷
We heard you! We want to bring Basecamp to Brazil in a big way too, so we're on the lookout for a full-time Country Manager to lead the movement.
You will be the face of Basecamp in Brazil. Your mission? Spread the word about Basecamp's simplicity, get teams excited about its features, and create a buzz that resonates throughout the local scene. We have hundreds of Basecamp customers and fans in Brazil today, but we'd like to see thousands more.
If you live in Brazil, are fluent in both English and Portuguese, and you're a spirited community enthusiast with a passion for Basecamp, we're eager to connect with you! Join us on a 6-month adventure that could extend beyond, with a compensation of $2000 USD/month, plus 30% commission on new Basecamp subscriptions.
We look forward to getting to know you!
Complete details on the job and how to apply here:
apply.workable.com/37signals…
Eric Magalhães retweeted
“You can’t avenge your grievances by squandering your potential. You can’t save talent or energy for better days. Whatever the situation, whatever the company, you should show up with your very best. Not for them, but for you.” world.hey.com/dhh/acting-you…
Eric Magalhães retweeted
"There's something inherently counterintuitive about this notion that people who feel overwhelmed by life, or lost in its endless possibilities, might not need a lighter load, but a more meaningful burden instead." world.hey.com/dhh/the-respon…
Eric Magalhães retweeted
> you buy so many domains, start so many side projects, but never seem to finish or launch any?
me:
Eric Magalhães retweeted
2021 was our biggest Black Friday Cyber Monday ever! Together with our friends at @GoogleCloud we achieved near-perfect uptime while averaging ~30TB/min of egress traffic across our infrastructure. That’s a massive ~43PB/day!
Here are some of the most interesting stats 🧵👇
Eric Magalhães retweeted
I'm looking for an #AWS #Certification Specialist to join my team here at @cloudacademy! If you love building content, have a passion for AWS, enjoy teaching others, and hold multiple AWS certs, I would love to hear from you!
Apply here: jobs.lever.co/cloud-academy/…
#cloud #train
Eric Magalhães retweeted
ALOU Pessoa Desenvolvedora com experiência em ✨ React Native ✨
Vaga para nível pleno, 100% remota e CLT com faixa entre 6k e 7k.
Quem interessar, só me dar um alou com o link do linkedinho na DM!
Rt da amizade ai meus anjo? 💙





























