john_lam Profile Banner
John Lam Profile
John Lam

@john_lam

Followers
5K
Following
14K
Media
628
Statuses
7K

Product @Microsoft, making AI with Python more awesome across the company, including Excel! Model manager.

Redmond, WA
Joined March 2008
Don't wanna be here? Send us removal request.
@john_lam
John Lam
10 months
Python in Excel is now generally available!. On August 19, 2019 I met @keyurp32 and Shaofeng Zhu from the Excel team for the first time. They showed me a hackathon project that they worked on to integrate Jupyter notebooks into the Excel task pane. Fast forward 5 years and what
Tweet media one
Tweet media two
4
30
164
@john_lam
John Lam
1 month
Here's the gist from above: Here's a link to the repo:.
0
0
1
@john_lam
John Lam
1 month
I love the @AnthropicAI share feature. But I use Claude Code all the time and I really want to share transcripts of what I've learned from coding / writing sessions. So I had my minions write a tool that shares your Claude Code session logs as @github gists:. $ uv tool install
Tweet media one
2
0
12
@john_lam
John Lam
2 months
RT @pierceboggan: we are excited to build in the open with the @code community :) .
Tweet media one
0
40
0
@john_lam
John Lam
2 months
Here's a side by side picture:
Tweet media one
0
0
1
@john_lam
John Lam
2 months
These are my detailed notes: I was seriously thinking about a Framework 14 but the price performance and known quality of this machine swayed me in the end. Share with us what I'm missing in my setup!.
Tweet card summary image
notes.iunknown.com
Dialing‑In My Developer MacBook Pro (2025 Edition) A diary‑style walkthrough of every tweak, install, and automation that transformed a refurbished laptop into my perfect daily driver. Hardware at a …
1
0
4
@john_lam
John Lam
2 months
I bought a new-to-me MBP 16 M2 Max to upgrade from my MBP 14 M1 Max. This is a quality of life upgrade - mostly wanted the bigger display and 50% more RAM doesn't hurt. Was surprised that keyboard feels and sounds better. But most fun of all, I set it up from scratch with lots of
Tweet media one
1
0
5
@john_lam
John Lam
5 months
A new benchmark emerges for agentic AI systems: factorio. You can watch Sonnet 3.5 complete tasks or build entire factories in factorio.
Tweet media one
0
0
5
@john_lam
John Lam
6 months
I've tried creating a few reports using ChatGPT Deep Research and while they all look impressive, the model is hobbled by its lack of access to the primary source literature. There are citation level hallucinations happening which undermines confidence in the outputs. On an area.
0
0
4
@john_lam
John Lam
6 months
@paulgauthier It's getting worse. This is a bread and butter coding with LLM prompt. Claude did a great job here as my backup AI.
Tweet media one
Tweet media two
0
0
1
@john_lam
John Lam
6 months
My recent usage pattern for o1 pro is to use it as an architect / spec writing model and to use sonnet as the coding model, which is what @paulgauthier does with Aider. This morning I had o1 pro refuse to give me the answer to my prompt 4 times in a row. This is the prompt and
Tweet media one
1
0
2
@john_lam
John Lam
7 months
Does anyone know if you can create a GPT and get it to use a specific model? As far as I can tell, GPTs seem to run on the default model - I'm assuming gpt-4o. I really just want to create a system prompt that I can select that tells the model to make minimal changes to the.
0
0
1
@john_lam
John Lam
10 months
Getting my money's worth talking to o1-preview. This thing is a spec writing monster.
Tweet media one
0
0
5
@john_lam
John Lam
10 months
Here's the launch announcement from the team with details on how to get started with the experience: There's a lot more work to be done here, and although I haven't been involved in the day to day of this project for over a year now, I continue to cheer.
Tweet card summary image
techcommunity.microsoft.com
 Python in Excel is now generally available for Windows users of Microsoft 365 Business and Enterprise. Last August, in partnership with Anaconda, we...
2
1
13
@john_lam
John Lam
11 months
Dear @cursor_ai : please make sure you never ever lose the prompt that I'm typing. I just lost 15 minutes of work typing in a prompt because of some bug. @AnthropicAI does a great job in their UI of never losing what I've typed in, regardless of what stupid thing I happen to do.
2
0
4
@john_lam
John Lam
1 year
This episode is exceeding expectations. It took me nearly 1 hour to listen to the first 5 minutes. It's that good. Make sure you go and READ THE TRANSCRIPT. There are links to all sorts of rabbit holes in it. I am so thankful that this exists in my world. 🙏.
@dwarkesh_sp
Dwarkesh Patel
1 year
Had so much fun chatting with my friends @TrentonBricken and @_sholtodouglas. No way to summarize it, except:. This is the best context dump out there on how LLMs are trained, what capabilities they're likely to soon have, and what exactly is going on inside them. You would be
0
0
3
@john_lam
John Lam
1 year
I don’t think I’ve looked forward to a podcast as much as this one.
@dwarkesh_sp
Dwarkesh Patel
1 year
.@TrentonBricken explains how we know LLMs are actually generalizing - aka they're not just stochastic parrots:. - Training models on code makes them better at reasoning in language. - Models fine tuned on math problems become better at entity detection. - We can just
0
1
35
@john_lam
John Lam
1 year
RT @vboykis: I wrote a very long post on what a machine learning artifact is, how files work, safetensors, pickle, and a lot more, as a way….
0
65
0
@john_lam
John Lam
1 year
The reading list to prepare for your upcoming @openai interviews. Seriously, this is like a syllabus for a graduate level course in how LLMs are built on the shoulders of giants. It's a love letter to all the incredible engineering that brought us to this moment today.
@itsclivetime
Clive Chan
1 year
Super long 🧵 incoming. I really enjoy looking at the full stack of computing - maximizing intelligence per picojoule. Been wanting to do a far-too-in-depth overview of the ENTIRE thing - from manufacturing a GPU to training a GPT :D. Here we go!
Tweet media one
1
0
10
@john_lam
John Lam
1 year
This is the way.
@leguilloux
Christian LeGuilloux
1 year
Not a fan of the giant book bag sized cases for the Vision Pro. I want to easily be able to take it with me as discreetly as possible while saving space. So far ski goggle cases do work.
Tweet media one
Tweet media two
Tweet media three
0
0
4