Andrew Gross

There have been times that I was using an agent to add features to my own OS library, where it referred me to use... my own open source library.

I can see companies wanting to expose a way for your agent to hook into their systems/data, but with a better interface than just some query APIs. They would have their own prompting, skills, agents etc, but would be "prompted" by your local agent. Maybe the A2A protocol has something like this.

Has anyone seen support for things like "remote skills" for Claude Code or similar? Its a bit different from MCP in that its not just calling a regular API endpoint. Im thinking something closer to how the Web Search tool works, where its really a mini remote agentic system returning results.

YouTube video by Adam Hennings Auto memory and MEMORY.md in Claude Code

Looks like at least 10 days ago www.youtube.com/watch?v=l3O4...

Manage Claude's memory - Claude Code Docs Learn how to manage Claude Code's memory across sessions with different memory locations and best practices.

code.claude.com/docs/en/memo...

Picture of the auto memory description text from https://code.claude.com/docs/en/memory

Picture of a claude code session where the agent decides to write to the MEMORY.md file for the project.

When did Claude Code auto memory start getting rolled out, cool as hell

SWE-rebench Leaderboard SWE-rebench: A Continuously Evolving and Decontaminated Benchmark for Software Engineering LLMs.

Have you been following model performance on SWE-Rebench to attempt to identify contamination? swe-rebench.com

When agents have metrics, output tests, and input data, they can iterate like crazy. Being able to generate this loop this approach for any problem would be a huge timesaver and remove a lot of mental effort.

For the next big jump for coding agents, I think the frontier labs are going to spend a lot of effort on making them good and creating their own problem specific harness for iterating.

I want to make the concept of cells, output, and state in an ipykernel legible to them as well for the same reason. Much easier to collaborate when the end goal isnt just a top to bottom script every time.

I have a sneaking suspicion time travel debuggers are going to come back in to vogue once they are made legible to agents.

Maybe the harness needs to start with setting a target iteration speed and using it as an optimization metric.

Getting agents to run experiments where you need to trade off feature optimization with running time optimization is tough. Agents don't experience linear time and don't have a "feeling" that they need to spend time optimizing to iterate faster.

There's no reason the "Movie" Toothbrushing song on the Yoto has to go that hard.

One of those cases where it would have been astounding to me if someone hadn't already investigated these problems deeply, I just didn't know how to find it.

I was doing a bit of work with taking disjoin subgraphs and wanting to separate some of the larger ones based on overlapping connections and got introduced to the Louvain method and bridging.

LLMs can be great for those cases (in coding at least) where you assume there is a body of work around a problem, but you don't know the terminology to find it. In the past you just had to Google and hope, or ask a coworker. Still possible to fail with LLMs but can be easier.

Github Stars are a very salient example of Goodhart's law right now.

I feel like a time travel debugger that lets you rewind state while making edits etc could be super helpful for them, given the right interface. Makes it easier to take actions that are easily reversible.

My kingdom for sparse mxfp4 kernels in FlashInfer for sm120

CC behaves differently with Shift-Enter in the terminal vs as a Cursor extension. Its maddening.

If if you let it iterate on local hardware before crafting Kaggle notebooks for submission, I bet a lot of the trick would be to get it to be able to run and monitor long async jobs, and make decisions about when to optimize training speed.

I want to figure out a harness so that I can run Claude Code against a Kaggle competition. I feel like it could pretty reliably score silver in many competitions, assuming it was given access to the forums and had a decent harness

ASCII/text flow chart with an arrow between nodes where the arrow character's shaft isnt quite aligned with the line character's height

New requirement for monospace fonts: When coding agents draw diagrams the arrow lines are aligned.

I used to think that typing speed beyond a certainly level had diminishing returns for programming. That said, it seems like there is a lot of value in being able to quickly convey prompts via typing, as voice transcription lacks some capabilities for things like linking files or code.

I remember a while back there was a lot of discourse around codebases that are "greppable" vs those that are expected to be navigated with an IDE. Seems like greppable ended up being the correct choice for coding agents (at leat for now).

Trying out a new thing: "This code dictated but not read" to absolve myself of all responsibility for LLM generated code, just like executives in the 60s would place all the blame for their bad ideas on their secretary.

github.com/andrewgross/...

Release PySpark Toolkit 0.5.0:

* support for S3 Presigned URLs in Pure Pyspark
* `fdtf` - A generic UDTF decorator that can wrap an arbitrary row level python function and append its results as columns to your dataframe. Cuz UDTFs are annoying to use.

One thing I noticed with a lot of AI blog spam is that they never seem to properly link things. I just went over an entire article about a tool I wanted to check out, and not once did it actually link to the tool.