Viewing profile — ford

I've been showing people chatjimmy for months - it's incredible. Both reasoning and tool use generation scale with TPS. Imagine 100x more reasoning on a model, or 100x parallel too…

I'll hijack this point to talk about robotics - I think we agree 1) It's not ideal for China to control production for _all_ of the things we use here. 2) What they already are goo…

Without the singularity, I think Frontier labs will offer intelligent model blends. They'll have their own versions of "cheap" models and be expert at using the appropriate amount …

Frontier models need to do everything for everyone. It's expected (though not often done) that smaller models fine-tuned on specific tasks can approach frontier performance on a sp…

In my experience everything that affects engineers affects an agent. Good abstractions, reasonably sized methods, good names, principled (intra & inter) service architecture, unit …

Related - "The Electric Slide" [0] from NotBoring (a techno-optimist news letter) talks about production of batteries and other parts of the "Electric Stack", and explains where th…

Basically MCP is little more than a brand name for "APIs LLM's can use". This means more services are creating APIs, because xyz company who's never been super tech forward doesn't…

> If you’re going to use an LLM to write me an email, I’d much rather you just send me the prompt; at least then I’d have an idea of what you actually meant to say.” I'm excited fo…

In a post Claude Code world that's the job of engineers - the engineering is designing good abstractions, scalable systems, and things that are easy to contribute to. This is what …

I think a more realistic model is not fully open source, but apps with extremely open/flexible APIs and data models that allow arbitrary front-ends (likely with a default one provi…

I'm really interested in how LLMs will enable more customizable, personal software. Our PMs & Designers are writing a lot of code now, and our engineers are spending time figuring …

The "open source" part is the wrapper on top (up to you if you believe that's meaningful here)

Good engineering has always been about minimizing the amount of effort it takes for someone to understand and modify your code. This is the motivation for good abstractions & inter…

I've found it really valuable to pair with people, sit at a computer together while they're driving and using AI. It's really interesting to see how other people prompt & use AI to…

Interesting thought (I think recently more than ever it's a good idea to question assumptions) - but IMO abstractions are important as ever. Maybe the smallest/most convenient pack…

I hear this often, but I think this discounts the fact that this was mostly true for the US/Western Europe at a time where they enjoyed unilateral super-powerism as a result of win…

Get a "Year-In-Review" for arbitrary data. Yirgachefe was borne out of 3 ideas: 1) Spotify Wrapped is delightful 2) In many countries, services are required by law to provide your …

Has it been many people's experience that big companies intentionally remove experienced engineers from your team to something unrelated, in the name of fungibility? I've surely se…

At the end of the day writing good code is rarely the "end" someone is shooting for. It's more research, more features, more experimentation, etc. Maybe hobby projects and library …

"Why are we trying to make Yahoo Search faster? I already am fine with my 2-3s wait time"

I think we take git for granted as software engineers Software engineering has decades of experience with proposing changes, merging them, staging them, deploying them, and rolling…

IMO the only concentration OpenAI has is brand. Anthropic & Gemini both have roughly equivalent models. This could change quickly since success compounds, but for now I am actually…

The argument sounds like he believes AI (+ robotics) will take jobs, and breaking up OpenAI could slow it down Historically the most productive countries are the most prosperous - …

Interestingly I've often seen this in Claude outputs, especially on long prompts. I've assumed this is because of Claude's XML-based instruction format, but this does make me wonde…