Playing with Fire in AI Coding
Skills:
AI Pair Programming80%
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all of my documentation for the community and all of that stuff, if I look through that, I'm not going to say like 80% is obsolete, but it's not it's definitely a majority is obsolete.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
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
Our 1st MLOps Meetup // Luke Marsden // MLOps Meetup #1
MLOps.community
Remote Collaboration as a Data Scientist
MLOps.community
MLOps Manifesto with Luke Marsden from Dotscience
MLOps.community
MLOps lifecycle description
MLOps.community
What Does Best in Class AI/ML Governance Look Like in Fin Services? // Charles Radclyffe // MLOps #2
MLOps.community
Life purpose and too many spreadsheets
MLOps.community
Explainability, Black boxes and EU white paper on reproducibility
MLOps.community
Hierarchy of Machine Learning Needs // Phil Winder // MLOps Meetup #3
MLOps.community
Automatically Retrain Machine Learning Models? Are best practices worth it?
MLOps.community
Building an MLOps Team? Key ideas to keep in mind
MLOps.community
Hierarchy of MLOps Needs
MLOps.community
Bare necessities for getting an ML model into production
MLOps.community
MLOps and Monitoring
MLOps.community
How Phil Winder got into Data Science and Software Engineering
MLOps.community
Provenance and Reproducibility in Machine Learning; what is it and why you need it?
MLOps.community
Friction Between Data Scientists and Software Engineers
MLOps.community
MLOps Problems in different size companies
MLOps.community
ML tooling in large companies
MLOps.community
ML Platforms - The build vs buy question
MLOps.community
ML Services Gateway at SurveyMonkey
MLOps.community
Message buses, Async and sync architecture
MLOps.community
MLOps #4: Shubhi Jain - Building an ML Platform @SurveyMonkey
MLOps.community
Hybrid Data Science Teams @SurveyMonkey
MLOps.community
How do you handle ML version control at SurveyMonkey
MLOps.community
Doing ML with Personal Information
MLOps.community
Evolution of the ML feature store @SurveyMonkey
MLOps.community
Developing a Machine Learning Feature Store
MLOps.community
Auto retrain ML models is not the question
MLOps.community
3 key parts to Machine Learning monitoring
MLOps.community
MLOps Meetup #6: Mid-Scale Production Feature Engineering with Dr. Venkata Pingali
MLOps.community
MLOps meetup #5 High Stakes ML: Active Failures, Latent Factors with Flavio Clesio
MLOps.community
MLOps: Airflow Pros and Cons
MLOps.community
Specific challenges in Machine Learning
MLOps.community
Current State Of Machine Learning
MLOps.community
Humans in the Loop are a defining factor in Machine Learning
MLOps.community
Learning from real life Machine Learning failures
MLOps.community
Survivorship Bias in machine learning tutorials
MLOps.community
Swiss Cheese model in Machine Learning
MLOps.community
Resume driven development in Machine learning & software engineering
MLOps.community
Who has the highest standards in ML?
MLOps.community
Venkata Pingali of Scribble Data Thoughts on the Current State of Machine Learning
MLOps.community
Dependable data and being able to Trust in your Data with Venkata Pengali of Scribble Data
MLOps.community
Speed, Trust, Evolution and Scale in MLOps
MLOps.community
More difficult transition for data scientists to become ML engineers
MLOps.community
How many models in prod til I need a dedicated ML platform?
MLOps.community
Deeper thinking from data scientists around platform blackholes
MLOps.community
Checkpointing, metadata, and confidence in your data
MLOps.community
Adjacent usecases and multistep feature engineering
MLOps.community
Standardization of Machine Learning tools like in Software Engineering with Venkata Pingali
MLOps.community
Reproducability flaws in end to end Machine Learning debugging
MLOps.community
3rd wave of data scientists
MLOps.community
MLOps meetup #7 Alex Spanos // TrueLayer 's MLOps Pipeline
MLOps.community
MLOps Meetup #8 Optimizing Your ML Workflow with Kubeflow 1.0
MLOps.community
Are Kubeflow and Airflow complementary?
MLOps.community
Why Kubeflow gained so much traction=open community
MLOps.community
Who decides the dirrection of Kubeflow
MLOps.community
What do Kubeflow and Arrikto do and how do they work together?
MLOps.community
Versioning your ML steps with Kubeflow
MLOps.community
Machine Learning Lifecycles//Perception vs Reality
MLOps.community
Kubeflow vs SageMaker in Machine Learning
MLOps.community
More on: AI Pair Programming
View skill →Related Reads
📰
📰
📰
📰
How to audit a Lovable, Bolt or v0 app before launch: Audit Vibe Coding by Inithouse
Dev.to AI
Amid hardware legal battle, OpenAI releases a $230 keyboard for Codex
TechCrunch AI
7 advanced Claude Code tips from 17 months of intense use
Dev.to · YK
Three handoff boundaries around a coding-agent workflow
Dev.to · Nekoautomata Miki
🎓
Tutor Explanation
DeepCamp AI