nana🦄

data scientist from a non tech background. code with coffee. share my learnings here #datascience #python #rstats #sql

[nana🦄 @ds_nana_]

AI Engineering Specialization

🚀 Intro to AI Engineering
🔓 Open-source AI Models
🧠 Learn Embeddings and Vector Databases
🤖 Learn AI Agents
💬 Learn OpenAI's Assistant API
📱 Build AI Apps with LangChain.js

https://t.co/a565bEYZJ6

[Samuel Wong @samuel_wong_]

Control the ideas, not the code

The 8-hour trade-off matters. Every hour spent reading code is an hour not spent on what's now the most valuable work: thinking about what the software should do, where it should go, QA, and new ideas.

https://t.co/OUhq7TH88t

[Samuel Wong @samuel_wong_]

Most enterprise failures aren't resource or tech problems — they're shared-model failures. The organisation doesn't understand its customers, can't translate technical knowledge to decision-makers, or has alignment that exists on paper but not in execution.

[Samuel Wong @samuel_wong_]

The real reason you think of the perfect thing to say an hour too late

https://t.co/PE7YdbMQVg

[Zynex @0xzynex]

One guy built Kimi K3 alone and made over $10,000,000 in a single month. Then he went on stage and told everyone exactly how he did it. Almost nobody pressed play.

The talk sits on YouTube, one engineer walking through the machinery under Kimi, the open models that quietly closed China's six-month gap to zero.

Minute 5 kills the data wall, he shows how a new optimizer turns 50 trillion tokens into the equivalent of 100 trillion, calls it "almost like magic," and means it.

Minute 17 is the part US labs would never publish, a linear attention design that beats full attention across short and long context while staying cheap at a million tokens.

Minute 23 is the swarm, reward functions for orchestrating up to 1,000 agents in parallel, including a term that catches sub-agents faking work they never finished.

Minute 25 he puts up a training curve, 30 trillion tokens with not a single loss spike, and calls it the most beautiful curve of his life.

Consultants charge $500 an hour to guess at this. The primary source is free and timestamped.

Everyone's arguing about the model. The blueprint has almost no views.

[Codez @0xCodez]

Andrew Ng just released a 1-hour course on building agentic knowledge Graphs from scratch:

• 00:00 - Introduction to agentic knowledge Graphs
• 03:07 - Construction of agentic Graphs
• 14:00 - Architecture of multi-agent systems
• 23:00 - Building agentic graphs with Google ADK
• 01:06:03 - Why Graphsare the future of agentic AI

Worth more than 10 articles on loop engineering.

Watch it today, then read how to become a graph engineer in the article below.

[Arena.ai @arena]

Exciting update: Kimi K3 has landed at #4 on the Agent Arena leaderboard, matching Claude Opus 4.8 and GPT-5.6 Sol. If Kimi K3's weights are released on schedule by July 27, it will become the #1 open-weight model.

This release marks a major leap in agentic performance over Kimi K2.7 Code (#23 to #4). Based on 8K+ live agentic sessions, Kimi K3 leads on confirmed task success rate (#1). It also posts a strong +20.6% on praise vs. complaint (#3). It currently lags the field in steerability (#14) and bash recovery (#17).

Agent Arena measures models on millions of real-world, long-horizon agentic tasks. Models get web search, filesystem, and terminal tools to complete complex workflows: writing code, creating slide decks, researching the web, building apps, and analyzing documents.

We use causal tracing methodology to measure a model's net improvement, which indicates how much it improves outcomes relative to the average model.

Here's a primer on the 5 signals:

User-satisfaction proxies
- Confirmed Success: an explicit "yes that worked" feedback from the user
- Praise vs. Complaint: implicit sentiment in users reactions
- Steerability: can the model course-correct when you push back?

Tool-use proxies
- Bash Recovery: how it recovers from CLI errors (primary signal for tool use)
- Tool Hallucination: does it call tools that don't exist

Below we break down how Kimi K3 scored across the 5 signals, drawn from tasks submitted by a global community of users.

Congrats @Kimi_Moonshot on another big milestone!

[fafa.👩🏻‍💻 @ds_fafa_]

Don’t type, just speak

The voice-to-text AI that turns speech into clear, polished writing in every app

Available on Mac, Windows, iPhone, and Android

https://t.co/S7CuHt2eBY

[Z @TaoMachina]

https://t.co/QGPUuYIqHr

[fafa.👩🏻‍💻 @ds_fafa_]

Google's AI Essentials course is the one I recommend to friends who ask "where do I start with AI?"

📚 5 courses, ~10 hours total
🤖 Covers prompting, responsible AI, productivity tools
⭐ 4.8 rating, 1.8M+ students

https://t.co/KkXyhN2Ktp #AI #MachineLearning #LearnAI

[Samuel Wong @samuel_wong_]

How Expedia Group Builds AI That Lasts at Scale

A framework for how to build, deploy, and evolve AI systems for impact and scale

https://t.co/FkGH0ci6e9

[Samuel Wong @samuel_wong_]

The Rosetta Stone For Business’s AI Failure Is Rewriting Every Job in Tech

https://t.co/gdO3n9qOxs

[wen👩🏻‍💻 @ds_wen_]

The Data Anarchy Tax: Why your team is firefighting 45% of the time

Results of the June 2026 pulse survey on organizations and data engineers

https://t.co/VMOV95YQJV

[Alex Prompter @alex_prompter]

Google Brain co-founder and Stanford professor, Andrew Ng, spent 149 minutes explaining how to prompt AI in 2026 better than any paid prompting course on the market.

This is what separates AI novices from power users:

1. "Think step by step" is dead advice.
This was standard prompting wisdom as recently as 2023. Andrew says the models have outgrown it.
"I no longer tell my AI model to think step by step. Instead, I'm more likely to just tell it to think hard. It knows what that means."
The most-repeated prompting tip on the internet became obsolete while most people kept using it.

2. Power users have empathy for the AI.
In an operational sense: they imagine being on the receiving end of their own prompt.
"If you could put yourself in the shoes of someone getting instructions from you, you can ask yourself, will they know enough about you to do a good job on the task you're assigning them?"
The best prompters anticipate what information the AI is missing.

3. Your biased questions get biased answers.
Andrew tested what happens when you frame a question with your preferred answer baked in. A Washington Post study he cites found ChatGPT agreed with users 10 times more often than it disagreed.
"If you give even a hint of what answer you're hoping for, there's a good chance the AI will just reflect back your preferences or your preconceptions."
Load a question with your preferred outcome and you're paying for a mirror and calling it analysis.

4. Never write the final text first.
Andrew Ng uses progressive outlining: outline first, critique the outline, iterate, expand to bullets, iterate again, and only then generate the final text.
"Editing the outline is very high leverage because you can change just a few words of the outline and this will result in an entire section of the article changing."
One edit at the outline level moves thousands of words. One edit at the sentence level moves one sentence.

5. Give feedback on options, not more instructions.
Instead of writing a longer prompt, Andrew asks AI for 3 to 5 options, then critiques them. His feedback becomes the context.
"One of the really good ways to figure out what additional context to give the AI is to give it feedback on the options it presents to you."
Your reaction to AI's output tells it more about what you want than another paragraph of instructions.

6. AI defaults to Reddit. Steer it to better sources.
Andrew cites data showing the most-cited website by AI models during web search is Reddit, followed by Wikipedia.
"If you don't steer the model in terms of what types of sources you prefer, there's a chance that it'll tend to pull text from whatever is most available rather than what's most reliable."
Ask for medical or financial information without specifying sources and you're getting Reddit-grade answers in professional packaging.

7. AI has jagged intelligence. No single model wins at everything.
Different models excel at different tasks, and which one leads changes with every new release. Andrew tests the same prompt across multiple models.
"There's some tasks where AI does poorer than human and some where it does much better than human and different AI models are jagged in different ways."
You get better results testing across providers and updating your intuitions about which tool fits which task.

Watch the full 149-minute course, then read the prompting playbook on GPT-5.6

[Rahul @sairahul1]

I don't prompt Claude Code anymore.

I have loops that prompt Claude Code for me.

My job is to write the loops.

This is the Boris Cherny method and it is unbelievably powerful.

Instead of:

→ writing a prompt
→ waiting for the answer
→ checking the output
→ fixing the mistakes
→ writing the next prompt

you design a system that does all of that itself.

The loop:

→ receives a goal
→ decides what to do next
→ chooses the right model
→ executes the work
→ checks the result
→ retries when something fails
→ stops only when the outcome is verified

You stop being the person operating the agent.

You become the engineer designing how the agent operates.

That is the shift happening right now:

Prompt engineering → Loop engineering

And once you understand it, you can build:

→ coding agents that work while you sleep
→ research systems that verify their own findings
→ software factories with parallel workers
→ recurring agents triggered by Slack, email or schedules
→ multi-model workflows that use the best model for each step
→ long-running systems with memory, state and real stop conditions

Most people still use AI one prompt at a time.

The best builders are creating systems that keep working after they leave.

I wrote the complete beginner's guide to getting started:

Loops & Harnesses: Why The Best AI Engineers Stopped Writing Prompts

It explains:

→ what a loop actually is
→ how every coding agent works underneath
→ the difference between a prompt and a job
→ planners, workers and independent verifiers
→ persistent state and memory
→ model routing
→ parallel execution
→ stop conditions
→ how Programs by Slate let you build your own loops in JavaScript

Everything you need to start loop engineering is below.

[Fiction @0xF1ction]

Kimi CEO Zhilin Yang:

"Every AI lab, like Claude, thinks the model is what matters most. That's wrong.

It's how you organize the people building it that wins. and that's about Kimi 3"

He spends 15 minutes tracing this back to a 1970s Intel chip nobody wanted to buy.

his one big idea: long context is basically the AI era's version of RAM

- the same jump from 128K to gigabytes, just compressed into a couple of years instead of forty

the real goal? "your biggest advantage, and perhaps your only advantage, is your organization"

watch & bookmark - then read the article on the org structure built for this bet ↓

[Khairallah AL-Awady @eng_khairallah1]

A senior Google engineer just dropped a 424-page doc called Agentic Design Patterns.

Every chapter is code-backed and covers the frontier of AI systems:

→ Prompt chaining, routing, memory
→ MCP and multi-agent coordination
→ Guardrails, reasoning, planning

This isn't a blog post. It's a curriculum. And it's free.

Read it this weekend.

That said, I wrote an article on how to build a multi-model AI team in 2026, which was already read by 4 million people.

It's a complete guide to agents, models, workflows, and playbook to run your business on autopilot.

The article is quoted below.

[Codez @0xCodez]

https://t.co/WVztfCUH4u

[Shubham Saboo @Saboo_Shubham_]

You can build, deploy and evaluate long-running AI agents using Claude Code, Codex or Antigravity.

Without writing a single line of code.

[wen👩🏻‍💻 @ds_wen_]

8 hours. That's all it takes to go from AI-curious to AI-capable. Google built the course.

📚 7 courses, beginner friendly
🤖 1M+ students, 4.8 rating
🛠️ Covers prompting, data analysis, productivity

https://t.co/ObJkHCcvcm #AI #MachineLearning #LearnAI

[Qwen Cloud @qwen_cloud]

Qwen3.8 is launching and going open-weight soon!🌐

With a massive 2.4T parameters, this model is continuously evolving. We believe it’s one of the most powerful model available today, compatible to leading frontier AI models , second only to Fable 5.

You don't have to wait to test it. Just now, the Qwen3.8-Max-Preview made its debut on Alibaba’s Token Plan, Qoder, and QoderWork. Be among the very first to try it out.

Can't wait to hear what you build. Stay tuned!

Token Plan ：
international：https://t.co/aP7QXKVNEX
China：https://t.co/vsq7elAnIr