oopsiedayysie

cursed shitposts for the avg techie in the AI gwangya 💀
too cheap to fire, too funny to lay off
my bias is continued employment
Skills UnFutured™

[oopsiedayysie @poetryrhythm_]

@jess_weng18 Thank you so much for replying! Love the memes and the way everything's structured! Would deffo attend more in other series of the webinar!

[oopsiedayysie @poetryrhythm_]

me: I should learn more about AI agents
LangChain: what if we made them DEEP
me: say less 💀

Agent Harnesses & Deep Agents 101 by LangChain @jess_weng18 resident meme extraordinare 💅, sent me down a rabbit hole today.

some things learnt along the way 🧵👇

[oopsiedayysie @poetryrhythm_]

6/ obviously, real implementations have more going on, but there's something delightfully un-mystical about looking at agents through the lens of control flow instead of ✨mysterious AI magic✨

less magic, more engineering. 💅

[oopsiedayysie @poetryrhythm_]

5/ honestly, one of my favourite takeaways was this mental model:

an agent is essentially a loop with access to tools ✍️

model decides → calls a tool (or doesn't) → observes the result → repeat until it's done or hits a stopping condition 🚏

[oopsiedayysie @poetryrhythm_]

4/ THE CONCEPT THAT REALLY GOT ME: Progressive Disclosure 👀

instead of dumping every tool's entire documentation and schema into context upfront, we start w more compact representation (eg tool names & description).

Expose more details only when needed. 📖 📖

[oopsiedayysie @poetryrhythm_]

3/ the journal entry demo was a neat way to see how this opinionated harness works live.

was really enriching to see tool calls! context management! how tools are decided.

me, watching the webinar: TECHNOLOGIA ✨⭐ 💖

[oopsiedayysie @poetryrhythm_]

2/ its open sauced agent harness, Deep Agents, handles planning, context management & multi-agent orchestration.

All batteries included!! 🔋

Rather than assembling everything from scratch, you get an opinionated harness.

[oopsiedayysie @poetryrhythm_]

1/ an agent is essentially a model calling tools in a loop until it completes a task and returns a result.

the model is one part of the equation; the harness is what gives it structure.

which is why I find LangChain's opinionated harness interesting

[Rohit Ghumare @ghumare64]

LeetCode, but for ML and AI.

We have LeetCode for DSA.

But where's the LeetCode for AI engineers?

This is not a 10-hour YouTube tutorial.

Your ordinary "Build an AI agent in 5 minutes" course.

I mean actual engineering challenges:

→ Implement backpropagation without PyTorch.

→ Build a tokenizer from scratch.

→ Write your own attention mechanism.

→ Build RAG without relying on frameworks.

→ Create an AI agent loop from scratch.

→ Learn how to evaluate and debug AI systems.

So I built something I wish existed when I started learning AI.

🔗 https://t.co/Q2FHIjaPI5

The idea is simple:

1. Understand the problem.
2. Learn the math behind it.
3. Implement it yourself.
4. Run the tests.
5. Actually understand what you built.

From ML fundamentals to transformers, LLMs, RAG, and AI agents.

✅ 100% open source
✅ Completely free
✅ No paywalls or signups
✅ Learn by building, not just watching

LeetCode taught a generation of engineers how to solve algorithms.

Maybe it's time we had something similar for building AI.

If you're serious about becoming an AI engineer, stop just calling APIs.

Start understanding what's behind them.

♻️ Repost this for someone learning AI.
💯 Completely Opensource and FREE with 71K stars

P.S. Would you use a daily LeetCode-style challenge for AI engineering?

[O Leitor @leitorpsique]

Minha terapeuta me disse:

"Se você é uma pessoa ansiosa, faça tudo por diversão. Vá a uma entrevista de emprego por diversão. Exercite-se por diversão. Comece um blog por diversão. A ansiedade se alimenta da importância. Não trate tudo como uma situação de vida ou morte."

E então tudo fez sentido.♟️

[Simular @SimularAI]

In another universe, you missed your kid's recital. Your mom's birthday dinner. That anniversary celebration with your person.

In this one, you have 𝐒𝐚𝐢.

The AI co-worker that does your computer work so you don't have to choose.