Niyi Royce

MetroSage || SWE || Ai & ML

[Niyi Royce @PyNiyi]

First Article of the year
https://t.co/Cr3P0u5Sg2

[expemilly @expemillyweb3]

A father asked Karpathy what to teach his 8-year-old before AGI. the answer erased 80% of the school curriculum.

math, physics, CS. everything else - tack it on later, if at all.

Most parents optimize their kid's schedule around grades and never notice they're funding the parts of the brain that stop mattering the day AGI ships.

grooves get harder to carve every year past 10.
the 8-year-old who learns physics processes reality differently at 30 than the one who didn't.
after AGI, information is a commodity. cognitive shape is not.

you're not choosing your kid's grades. you're choosing the brain they'll carry for the next 60 years.

16 minutes. free. changed what i'd tell any parent asking about school today.

watch it, then read below:

[Niyi Royce @PyNiyi]

Emotional intelligence
-Daniel Goleman
Is such a great book

[Niyi Royce @PyNiyi]

@cyrilXBT This is great.

[CyrilXBT @cyrilXBT]

A MIT professor gave a 1-hour lecture in 2019 that has 18 million views.

He died 5 months after recording it.

It was his final gift to the world.

Patrick Winston taught at MIT for 50 years.

The smartest engineers on earth sat in his classroom.

And he spent his last lecture teaching them the one skill their degrees never covered.

How to speak.

15 lessons that will change how you communicate forever:

Never open with a joke. Your audience is not ready to laugh yet. Open with a promise of what they will know by the end.

Your ideas are like your children. You are too close to them. What is obvious to you is invisible to everyone else. Explain the obvious.

The 5-minute rule: the first 5 minutes of any talk determine whether people will listen for the next 55. Spend more time on your opening than anything else.

Repeat your most important idea 3 times in 3 different ways. Once is never enough.

Build a fence around your idea. Tell people what it is NOT before you tell them what it IS.

Verbal punctuation. Pause. Let the idea land before moving to the next one.

Ask questions nobody will answer. Then wait 7 seconds. The silence is not awkward. It is processing.

Never read your slides. Your audience can read. They cannot listen and read simultaneously.

Use the board not the slides. Writing forces you to slow down. Slowing down forces clarity.

Inspire before you inform. Nobody learns from someone they are not inspired by.

End with a contribution not a summary. Tell them what you gave them. Not what you said.

Never say thank you at the end. It is weak. End with something that lands.

Stories make ideas stick. Data makes ideas understood. You need both. In that order.

The quality of your communication determines the quality of your ideas in the eyes of the world. Not the ideas themselves.

Practice is not preparation. Practice IS the skill.

Patrick Winston understood something most people spend their entire careers missing.

Your ideas are only as powerful as your ability to transfer them into someone else's mind.

You can be the smartest person in the room and be completely invisible.

Or you can master communication and make average ideas feel like breakthroughs.

He chose to spend his last lecture teaching this.

Watch it tonight.

Bookmark this first.

Follow @cyrilXBT for more lessons from the people who built the future.

[Byt3Mage @i_alayaki]

The compiler-generated bytecode for my VM is as optimal as if you hand-coded it. Zero movs

With only 256 registers per function, register allocation has to be very conservative and efficient. And with static typing (and no tagging), it's as fast as it gets.

She's beautiful🥹

[vicki @vboykis]

Recommended reading if you feel you’ve lost your way , or if you’re having more fun than ever

[tetsuo @tetsuoai]

Attention is a lookup. Each token builds a query, compares it against every key in the sequence, and pulls value vectors weighted by the match. Stack that 96 layers deep and you get a frontier model.

Video covers the full pipeline: Q/K/V, attention scores, encoder blocks.

[Deep-ML @real_deep_ml]

We just added interview prep for two Jane Street roles:
- Machine Learning Engineer
- Quantitative Researcher.

The prep walks you through what to study for each round, what to expect in the interview process, and which projects you can build for your resume to increase your odds of getting the interview in the first place.

[Andrew Ng @AndrewYNg]

“Loop engineering” is a hot buzzphrase after mentions of it by Boris Cherny (Claude Code’s creator) and Peter Steinberger (OpenClaw's creator) went viral on social media. Loops are now a key part of how we get AI agents to iterate at length to build software. In this letter, I’d like to share my 3 key loops, shown in the image below, for building 0-to-1 products. These loops guide not just how I build software, but also how I decide what software to build.

Agentic coding loop: Given a product specification and optionally a set of evals (that is, a dataset against which to measure performance), we can have an AI agent write code, test its work, and keep iterating until the code is bug-free and meets its specification. This idea of closing the loop took off around the end of last year, and it has been a game changer in enabling coding agents to work longer productively without human intervention. For example, over the weekend, I was building an app for my daughter to practice typing, and my coding agent could easily work for around an hour, using a web browser to check what it had built multiple times before getting back to me, without needing my intervention.

The engineering loop executes quickly. Every few minutes, the coding agent might build and test a new version of the software. I hear frequently from developers who are finding new ways to engineer more effective engineering loops. This is an active area of invention!

Developer feedback loop: In this loop, a developer examines the current product and steers the coding agent to improve it. Last year, a lot of developers (including me) were acting as the QA (quality assurance) function for our coding agents, manually finding bugs and then asking the agent to fix them. But with coding agents much more able to test their own code, the amount of time we need to spend on this function has decreased significantly. This allows us to make higher-level product decisions, such as what key features to offer, where the UI needs improvement, and so on.

The developer-feedback loop operates over time intervals between tens of minutes and hours — that's how frequently a developer might review a product and give feedback. In the case of the typing app, I changed my mind a few times about the visual design, what cat costumes she can unlock as she learns (she loves cats), and the user flow for a grown-up to log in and steer the child's learning experience.

When a developer has a clear vision for what to build, it is still a lot of work to translate that vision into a specification for a coding agent to implement. Further, after the developer has seen an implementation, they might update (or perhaps clarify) the spec to steer it toward what they want. If you find that the system repeatedly runs into certain problems, building a set of evals for the agent becomes useful.

AI-native teams are increasingly using AI to help shape product direction, for example, automating the gathering and analysis of usage data, summarizing written and verbal customer feedback, or carrying out competitive analysis. However, for pretty much all the products I’m involved in, I see humans as having a significant context advantage over current AI systems — we know a lot more than the AI system about the users and the context the product has to operate in — and thus humans play a critical role. Many people describe this human contribution as “taste,” but I prefer to think of it as humans having a context advantage, since that gives us a clearer path to helping AI systems get better. This also speaks to why this step can’t be automated: So long as the human knows something the AI does not, human-in-the-loop is needed to to inject that knowledge into the system.

External feedback loop: This includes a wide range of tactics like asking a few friends for feedback, launching to alpha testers, or putting the code into production with A/B testing. These tactics are usually slow, rarely taking less than hours and sometimes taking days or even weeks. This data informs the developer vision, which in turn continues to drive the detailed product spec, which in turn drives the coding agent.

With coding agents speeding up software development, more engineers are starting to play a partial product management role. For many engineers who are growing into this role, the hardest part is shaping the product vision and striking a balance between building (bridging the gap between vision and spec) and getting user feedback to evolve the vision. It is important to do both!

I will write more about how to do this in future posts, but for now, I find it encouraging that engineers are playing an expanded role (just as product managers and designers now do more engineering).

[Original text: The Batch]

[Niyi Royce @PyNiyi]

Greatness is within

[Freya Lawson @Freyabuilds]

this is f*cking gold

Andrej Karpathy joined Anthropic five weeks ago.

A friend on his team just showed me the exact Claude.md file he actually uses.

I dropped it into my setup. The very first response was different.

Not slightly different. Completely different.

Claude stopped giving generic answers and started working exactly the way I think.

Bookmark it before it gets lost in your feed.

Read it now, then check the article below.

[Max @minordissent]

To be highly intelligent is to be selected as one of the few human beings capable of solving actually hard problems. This makes you basically a shepard of humanity, the masses your retarded dependent children. To abdicate this duty and become a “just enjoy life” maxxing free-rider is an affront to God. And he will punish you accordingly.

[Niyi Royce @PyNiyi]

@Franc0Fernand0 @shrav_10 Spot on

[Niyi Royce @PyNiyi]

@JLarky For reference, here's my playbook : https://t.co/s2eLOsdTGM

[Niyi Royce @PyNiyi]

@JLarky One workflow that's worked well for me in agentic coding is getting the LLM to reason about intent first, before writing a single line of code.

[Niyi Royce @PyNiyi]

@JLarky LLMs benefit from the same principles you'd apply when building a compiler: explicit inputs, well-defined workflows, deterministic stages, and clear contracts.

The more structured your reasoning pipeline, the better the model performs.

[Niyi Royce @PyNiyi]

@JLarky You can take it further by defining pre-hooks and post-hooks for every prompt sequence,validation before generation, verification after generation.

At the end of the day, this is less about prompt engineering and more about domain expertise

[Niyi Royce @PyNiyi]

@JLarky When an existing codebase already has clear conventions, architecture docs, style guides, and project structure, the model has far more context to reason. That dramatically reduces boilerplate, lowers hallucination rates, and produces code that's consistent with the system.

[Niyi Royce @PyNiyi]

@JLarky Instead of prompting for an implementation, prompt for:

Intent
Requirements
RPO/spec structure
Architecture
Constraints
Edge cases

[will brown @willccbb]

one of the craziest bets of prime intellect was swinging for everything from the start. we operate datacenters, we build our own trainers and envs and sandboxes, we have a platform for large enterprises and solo devs, we train huge models, we publish. ~40 people btw. hiring, DM