Playing with Fire in AI Coding

MLOps.community · Intermediate ·💻 AI-Assisted Coding ·10mo ago

Key Takeaways

The video discusses the importance of documentation in AI coding, the challenges of keeping it up-to-date, and the role of humans and AI in creating high-quality documentation. It highlights the need for humans to clarify their thinking, capture key insights, and leave knowledge nuggets for future reference, while AI can assist in writing and providing context.

Full Transcript

It can be very scary to use a coding agent because if you are an engineer and you're asked to make a change and you don't know how to make the change and then you ask an agent to do it, it's you know really you know >> with fire. >> Yeah, you're playing with fire. And so kind of a prerequisite to making changes is generally understanding the system like what is the impact of your changes and so that also is kind of where um documentation or just knowledge um becomes more important. But for humans just like understand like what is the impact of this change? Where should it change? So then you can actually like review um and be a sort of co-pilot for the coding agent. So incredible about the ways that they consume information because that lines up with everything. And it it's funny how making the link so giving them almost like grounding them in code gives them a better chance of success. and then throwing as much information at them as possible really lines up with yeah like give all the context you can and they can sus out what is actually important for them. Now this feels like a perfect segue into the quality of knowledge and how you're thinking about making knowledge higher quality in general because as I was mentioning before we hit record again was I love documentation. I'm a huge fan of trying to write. I think a writing helps you clarify your thoughts. It helps you get down the most important stuff that you want to then take forward. But what I've noticed is you have to constantly be updating docs as things change. And you also have to recognize that a lot of things are just going to not be relevant after a certain amount of time and you kind of need to know which ones are relevant in those moments of time. So, how does that knowing that like if I look through my notion, which is where I keep all of my documentation for the community and all of that stuff, if I look through that, I'm not going to say like 80% is obsolete, but it's not it's definitely a majority is obsolete. And that's because like every podcast that we've had for the last 5 years, you know, I have a notion page on them. I'm not using those anymore, but they're there and maybe they can be referenced. Maybe it can be something. Or there's like strategy documents that I've written up in 2022 and those are not relevant. Or like reflection documents. All of this stuff feels like it would muddy up the waters on documentation. And if you want the highest quality documentation for your business, how do you think about like keeping it high quality? So I think you're not alone in that you know 80% of your notion is stale. Uh I think that's probably the norm and you know you know what's the the saying is like the instant you write docs they're stale. And I think that's really true. I think the way we think about it and I think generally is like a good framing is you want to like lean on humans on for what they're good at and then AI for what they're good at. And humans are really good about knowing what is actually important. Um so I think like you know writing is still a very useful and important exercise of like um you usually as like an expert on some do topic or having just done research like know what are the things that someone in the future need to know and sort of like maybe the story that you want to tell around it and you know AI can help you write but at the end of the day like you know your the nuggets of you know expertise or like knowledge that you as the expert like put into it like That's what's really really important to have the human in the loop behind. [Music]

Original Description

Ever opened your Notion and realized 80% of it is outdated? Devin Stein has too—and he says that’s the norm. Documentation goes stale the moment you write it, but that doesn’t mean it isn’t valuable. Writing helps you clarify your thinking, capture key insights, and leave “knowledge nuggets” for the future. The challenge is keeping it high-quality while everything around you changes. Devin’s take: 🔥 Using coding agents without context is like playing with fire. ✨ Humans are great at knowing what’s actually important. 🤖 AI is great at surfacing and organizing context. Together, they make knowledge *eventually consistent*. Full episode here: 🎧 **Knowledge is Eventually Consistent // Devin Stein // MLOps Podcast #335** https://home.mlops.community/home/videos/knowledge-is-eventually-consistent
Watch on YouTube ↗ (saves to browser)
Sign in to unlock AI tutor explanation · ⚡30

Playlist

Uploads from MLOps.community · MLOps.community · 0 of 60

← Previous Next →
1 Our 1st MLOps Meetup // Luke Marsden // MLOps Meetup #1
Our 1st MLOps Meetup // Luke Marsden // MLOps Meetup #1
MLOps.community
2 Remote Collaboration as a Data Scientist
Remote Collaboration as a Data Scientist
MLOps.community
3 MLOps Manifesto with Luke Marsden from Dotscience
MLOps Manifesto with Luke Marsden from Dotscience
MLOps.community
4 MLOps lifecycle description
MLOps lifecycle description
MLOps.community
5 What Does Best in Class AI/ML Governance Look Like in Fin Services? // Charles Radclyffe // MLOps #2
What Does Best in Class AI/ML Governance Look Like in Fin Services? // Charles Radclyffe // MLOps #2
MLOps.community
6 Life purpose and too many spreadsheets
Life purpose and too many spreadsheets
MLOps.community
7 Explainability, Black boxes and EU white paper on reproducibility
Explainability, Black boxes and EU white paper on reproducibility
MLOps.community
8 Hierarchy of Machine Learning Needs // Phil Winder // MLOps Meetup #3
Hierarchy of Machine Learning Needs // Phil Winder // MLOps Meetup #3
MLOps.community
9 Automatically Retrain Machine Learning Models? Are best practices worth it?
Automatically Retrain Machine Learning Models? Are best practices worth it?
MLOps.community
10 Building an MLOps Team? Key ideas to keep in mind
Building an MLOps Team? Key ideas to keep in mind
MLOps.community
11 Hierarchy of MLOps Needs
Hierarchy of MLOps Needs
MLOps.community
12 Bare necessities for getting an ML model into production
Bare necessities for getting an ML model into production
MLOps.community
13 MLOps and Monitoring
MLOps and Monitoring
MLOps.community
14 How Phil Winder got into Data Science and Software Engineering
How Phil Winder got into Data Science and Software Engineering
MLOps.community
15 Provenance and Reproducibility in Machine Learning; what is it and why you need it?
Provenance and Reproducibility in Machine Learning; what is it and why you need it?
MLOps.community
16 Friction Between Data Scientists and Software Engineers
Friction Between Data Scientists and Software Engineers
MLOps.community
17 MLOps Problems in different size companies
MLOps Problems in different size companies
MLOps.community
18 ML tooling in large companies
ML tooling in large companies
MLOps.community
19 ML Platforms - The build vs buy question
ML Platforms - The build vs buy question
MLOps.community
20 ML Services Gateway at SurveyMonkey
ML Services Gateway at SurveyMonkey
MLOps.community
21 Message buses, Async and sync architecture
Message buses, Async and sync architecture
MLOps.community
22 MLOps #4: Shubhi Jain - Building an ML Platform @SurveyMonkey
MLOps #4: Shubhi Jain - Building an ML Platform @SurveyMonkey
MLOps.community
23 Hybrid Data Science Teams @SurveyMonkey
Hybrid Data Science Teams @SurveyMonkey
MLOps.community
24 How do you handle ML version control at SurveyMonkey
How do you handle ML version control at SurveyMonkey
MLOps.community
25 Doing ML with Personal Information
Doing ML with Personal Information
MLOps.community
26 Evolution of the ML feature store @SurveyMonkey
Evolution of the ML feature store @SurveyMonkey
MLOps.community
27 Developing a Machine Learning Feature Store
Developing a Machine Learning Feature Store
MLOps.community
28 Auto retrain ML models is not the question
Auto retrain ML models is not the question
MLOps.community
29 3 key parts to Machine Learning monitoring
3 key parts to Machine Learning monitoring
MLOps.community
30 MLOps Meetup #6: Mid-Scale Production Feature Engineering with Dr. Venkata Pingali
MLOps Meetup #6: Mid-Scale Production Feature Engineering with Dr. Venkata Pingali
MLOps.community
31 MLOps meetup #5 High Stakes ML: Active Failures, Latent Factors with Flavio Clesio
MLOps meetup #5 High Stakes ML: Active Failures, Latent Factors with Flavio Clesio
MLOps.community
32 MLOps: Airflow Pros and Cons
MLOps: Airflow Pros and Cons
MLOps.community
33 Specific challenges in Machine Learning
Specific challenges in Machine Learning
MLOps.community
34 Current State Of Machine Learning
Current State Of Machine Learning
MLOps.community
35 Humans in the Loop are a defining factor in Machine Learning
Humans in the Loop are a defining factor in Machine Learning
MLOps.community
36 Learning from real life Machine Learning failures
Learning from real life Machine Learning failures
MLOps.community
37 Survivorship Bias in machine learning tutorials
Survivorship Bias in machine learning tutorials
MLOps.community
38 Swiss Cheese model in Machine Learning
Swiss Cheese model in Machine Learning
MLOps.community
39 Resume driven development in Machine learning & software engineering
Resume driven development in Machine learning & software engineering
MLOps.community
40 Who has the highest standards in ML?
Who has the highest standards in ML?
MLOps.community
41 Venkata Pingali of Scribble Data Thoughts on the Current State of Machine Learning
Venkata Pingali of Scribble Data Thoughts on the Current State of Machine Learning
MLOps.community
42 Dependable data and being able to Trust in your Data with Venkata Pengali of Scribble Data
Dependable data and being able to Trust in your Data with Venkata Pengali of Scribble Data
MLOps.community
43 Speed, Trust, Evolution and Scale in MLOps
Speed, Trust, Evolution and Scale in MLOps
MLOps.community
44 More difficult transition for data scientists to become ML engineers
More difficult transition for data scientists to become ML engineers
MLOps.community
45 How many models in prod til I need a dedicated ML platform?
How many models in prod til I need a dedicated ML platform?
MLOps.community
46 Deeper thinking from data scientists around platform blackholes
Deeper thinking from data scientists around platform blackholes
MLOps.community
47 Checkpointing, metadata, and confidence in your data
Checkpointing, metadata, and confidence in your data
MLOps.community
48 Adjacent usecases and multistep feature engineering
Adjacent usecases and multistep feature engineering
MLOps.community
49 Standardization of Machine Learning tools like in Software Engineering with Venkata Pingali
Standardization of Machine Learning tools like in Software Engineering with Venkata Pingali
MLOps.community
50 Reproducability flaws in end to end Machine Learning debugging
Reproducability flaws in end to end Machine Learning debugging
MLOps.community
51 3rd wave of data scientists
3rd wave of data scientists
MLOps.community
52 MLOps meetup #7 Alex Spanos // TrueLayer 's MLOps Pipeline
MLOps meetup #7 Alex Spanos // TrueLayer 's MLOps Pipeline
MLOps.community
53 MLOps Meetup #8 Optimizing Your ML Workflow with Kubeflow 1.0
MLOps Meetup #8 Optimizing Your ML Workflow with Kubeflow 1.0
MLOps.community
54 Are Kubeflow and Airflow complementary?
Are Kubeflow and Airflow complementary?
MLOps.community
55 Why Kubeflow gained so much traction=open community
Why Kubeflow gained so much traction=open community
MLOps.community
56 Who decides the dirrection of Kubeflow
Who decides the dirrection of Kubeflow
MLOps.community
57 What do Kubeflow and Arrikto do and how do they work together?
What do Kubeflow and Arrikto do and how do they work together?
MLOps.community
58 Versioning your ML steps with Kubeflow
Versioning your ML steps with Kubeflow
MLOps.community
59 Machine Learning Lifecycles//Perception vs Reality
Machine Learning Lifecycles//Perception vs Reality
MLOps.community
60 Kubeflow vs SageMaker in Machine Learning
Kubeflow vs SageMaker in Machine Learning
MLOps.community

The video teaches the importance of documentation in AI coding, how to keep it up-to-date, and the role of humans and AI in creating high-quality documentation. It highlights the need for humans to clarify their thinking and capture key insights, while AI can assist in writing and providing context. By watching this video, viewers can learn how to effectively collaborate with coding agents and create high-quality documentation for their AI projects.

Key Takeaways
  1. Assess the current state of documentation for AI projects
  2. Identify areas for improvement and update documentation regularly
  3. Use AI as a co-pilot for coding and review AI-generated code
  4. Collaborate with coding agents to improve code quality
  5. Provide context and information to AI agents to improve their performance
💡 Humans are good at knowing what is actually important, and AI is good at providing context and information, so collaboration between humans and AI is essential for creating high-quality documentation.

Related Reads

📰
How to audit a Lovable, Bolt or v0 app before launch: Audit Vibe Coding by Inithouse
Learn to audit AI-generated apps like Lovable, Bolt, or v0 before launch to ensure they are secure and accessible
Dev.to AI
📰
Amid hardware legal battle, OpenAI releases a $230 keyboard for Codex
OpenAI releases a $230 keyboard for Codex amid a legal battle with Apple, learn how to leverage AI-powered coding tools and keyboards
TechCrunch AI
📰
7 advanced Claude Code tips from 17 months of intense use
Master advanced Claude Code techniques after 17 months of intense use to boost coding productivity
Dev.to · YK
📰
Three handoff boundaries around a coding-agent workflow
Learn to implement three handoff boundaries for a coding-agent workflow to ensure safe and efficient collaboration between humans and AI agents
Dev.to · Nekoautomata Miki
Up next
How to Create ONE PAGE Website using Claude AI (FREE & FAST)
Quick Tips - Web Desiign & Ai Tools
Watch →