zerojames

I love your website design, cpach. There are many cool links to click, too!

> my main purpose is to record my thoughts/ideas/experiences

Mine too!

Share your blog! I was literally writing about starting blogs 10 minutes ago (https://jamesg.blog/2024/01/06/the-web-is-yours/).

You can take your blog in whatever direction you want; you can write technical, professional posts, or write about anything else that interests you! No matter what you write about, there is probably someone out there who will be interested in what you have to say. If you share your blog I'll take a look. Now you already have a prospective reader :)

I am unsure the extent to which my personal blog was weighed when I applied for my current job, but generally having a technical blog is a plus. Being able to point to blog posts where you have at least one post on a project you worked on, a bug you fixed, things you have learned, how you solved a problem, etc. is incredibly valuable.

How we validate each response varies depending on the test, but generally we are looking for exact equality. This is useful because we can measure whether GPT-4 with Vision does or does not, with full accuracy, solve a problem. This is a category of common questions related to LMM performance.

The model being "almost correct" is a valuable condition for us to document. I'll note the idea of having an "Almost" state that shows in yellow if the response for a given day is close to, but not exactly, right. There are use cases where this is valuable information. For instance, an OCR response being 90% correct may be tolerable if you can run error checking to reconstruct the response (i.e. the case for ISBNs or serial numbers constructed in a certain way).

I recommend writing on your personal website! You will own your content and you can design the site and the user experience as you want.

Search is only one method through which someone can find what you write. I read a lot of personal websites that I found out about through friends, HN, and links from one site to another.

I am continually cognizant of how many personal blogs I encounter that help me learn something I am curious about or solve a niche bug with which I have been struggling. I like to publish things I learn as I go to contribute back; if someone finds what I have written useful, I'm thrilled!

One thing at which I am trying to improve is to document personal coding challenges _as I go_ rather than documenting a finished state. I find it easier to write about something after it is complete rather than in the "messy middle" when I'm trying to come toward a solution. I'm trying to reflect more on the notion that documenting paths that one explored that went nowhere may be equally as valuable as the paths that went somewhere.

I am an active user of Monocle (https://monocle.p3k.io), a social reader UI (which is connected to a feed polling service, in my case Aperture). I can click "like" or "reply" and the reaction is published on my personal website. A notification is sent to the recipient via Webmention that I liked/commented on their content (if their site supports webmention).

Thank you!

I spent time over the holidays working on edits, document order, formatting, etc., but I did not consider that "this post" was confusing since everything would be concatenated into a book. This is excellent feedback. I appreciate it!

Good point! It has been years since I last published a PDF on the web (HTML has my heart :D) -- this is helpful information to have! I can set up a redirect to a more descriptive URL.

I started writing about READMEs a while ago but my writing on the topic is not polished. If you are curious, my notes are here: https://capjamesg.github.io/python-packaging-best-practices/...

Happy to elaborate as necessary!

Thank you so much! I sincerely appreciate it!

> minimum viable code snippets

When I read documentation, I love when there are a few snippets I can read, copy-paste, and adapt to work with my code, particularly when a library is vast. Google does a good job of this, for example. In some of the cloud code snippets, you can copy-paste and they _just work_ (after you have authenticated).

> e.g. without imports, so you can't just run it as is.

I think it only makes sense to omit imports when a guide is walking through a single file (i.e. step-by-step), or future code snippets are a continuation of existing code. I also love it when guides specifically call out what package(s) I need to install.

Thank you! The syntax was inspired by Mathematica. I like the idea of using verbs to describe actions that are built into the language.

I had three main motivations for the project:

   1. Learn how to write a programming language;
   2. Explore the extent to which I could use a new language to help people learn computer vision, and;
   3. Learn more about computer vision.

The goal was to create a language that is more suited for beginners, with the intent that you can build a vocabulary you can use to describe computer vision and programming more broadly. The extent to which this is/was achieved is unknown since I'd need to do a lot more work to get it in the hands of new developers, but nonetheless I enjoyed thinking and reading about this.

> sooner or later you find yourself having to implement all the features of a modern high level language

A friend told me "but James, you can just use Python for this!" While that is true, there is definitely something to be said about _how_ you do something, as well as what you can and can't do. How a language works and what it can and can't do is in the hands of the language designer; these decisions are technically interesting to make.

Roboflow | Open Source Software Engineer, Web Designer / Developer, and more. | Full-time (Remote, SF, NYC) | https://roboflow.com/careers?ref=whoishiring0124

Roboflow is the fastest way to use computer vision in production. We help developers give their software the sense of sight. Our end-to-end platform[1] provides tooling for image collection, annotation, dataset exploration and curation, training, and deployment.

Over 250k engineers (including engineers from 2/3 Fortune 100 companies) build with Roboflow. We now host the largest collection of open source computer vision datasets and pre-trained models[2]. We are pushing forward the CV ecosystem with open source projects like Autodistill[3] and Supervision[4]. And we've built one of the most comprehensive resources for software engineers to learn to use computer vision with our popular blog[5] and YouTube channel[6].

We have several openings available but are primarily looking for strong technical generalists who want to help us democratize computer vision and like to wear many hats and have an outsized impact. Our engineering culture is built on a foundation of autonomy & we don't consider an engineer fully ramped until they can "choose their own loss function". At Roboflow, engineers aren't just responsible for building things but also for helping us figure out what we should build next. We're builders & problem solvers; not just coders. (For this reason we also especially love hiring past and future founders.)

We're currently hiring full-stack engineers for our ML and web platform teams, a web developer to bridge our product and marketing teams, several technical roles on the sales & field engineering teams, and our first applied machine learning researcher to help push forward the state of the art in computer vision.

[1]: https://roboflow.com/?ref=whoishiring0124

[2]: https://roboflow.com/universe?ref=whoishiring0124

[3]: https://github.com/autodistill/autodistill

[4]: https://github.com/roboflow/supervision

[5]: https://blog.roboflow.com/?ref=whoishiring0124

[6]: https://www.youtube.com/@Roboflow

Keep making cool things :)

> other than "it's just for me" — and that's probably good enough

I love this sentiment, and I'm glad you made this! I love reading different syntaxes for programming languages. You should be proud. Writing programming languages is _hard_.

I made an abstract programming language in 2023 (https://visionscript.dev). The language still needs _a lot_ of work and some bugs fixed. With that said, I had so much fun making it.

Happy new year, everyone!

Thank you dang for all the work you do to help maintain HN!

I have been building a word game and I came across these datasets. I find the range of words delightfully quirky, and something that may be useful in a game.

One thing I noticed about count_1w.txt is there are brand names like Starbucks in there.

My first experience with static site generation was with. Jekyll was delightful. I spent almost all of my time writing HTML and CSS, interspersed with Liquid where necessary. I didn't need to think about the engine behind the system. Jekyll _just worked_.

I ran into some growing pains when I wanted to add more extensions so I used the opportunity to build my own site generator. My generator was greatly inspired by Preston-Werner. His blog post was a great read on the why behind designs that I am (and many others using many mainstream static generators are) using every day.

I am regularly behind the trends :)

> We don’t require an email, login, or payment to use our terminal (it is free as in speech and as in beer).

This excites me. I was excited to try a new terminal last week and I had to sign up. I wouldn't do that, so I went back to the default macOS one. I'll give Wave Term a try!

It's clear you have put a lot of thought into the product and its delivery after reading your message above -- it is noted and appreciated!

This is amazing!

The "Intentionally add images to your documentation" heading in the article summarizes my position on this topic well.

Screenshots can be a useful learning aid. For example, if you are guiding someone through a complex product action which involves several steps, a series of screenshots can be helpful.

If you are writing a tutorial, having a screenshot of output (i.e. a web page, an image, a UI) is useful for readers so they can visually match what they have to the result you have. (This is not applicable for text-only outputs).

All: I sincerely appreciate the time spent sharing feedback. Your notes and comments are helpful and give me tools to be a better writer .

Regarding the screenshots, I am not a fan of this approach. We adopted it because of the early trend to share ChatGPT screenshots, and to ensure people could see the origin of our prompting (the web interface).

I will start a discussion about screenshots with the team. This can be better.

I will also discuss the layout, too. Machine learning and AI is difficult enough. To the extent to which we can focus attention on the most important part of the page — the content — we should.

Thank you again for your notes! I appreciate it.

Thank you! We took screenshots of our tests, which you can download here: https://media.roboflow.com/lmms-tests.zip

(Gemini ones are not in there yet.)

Readability is important to me. What can I do better? I always appreciate feedback.

After some prompting, I got a response. It identified the dog in the image, but the dog was in the center point. I gave Gemini the Home Alone image and it couldn't identify the Christmas tree, returning invalid coordinates.

The post has been updated accordingly.

I have experienced intermittent performance with the web interface.

I am really curious about the object detection performance so I’ll be rerunning that test regularly.

Yes indeed!

P.S. I love https://technicalwriting.tools.

Thank you! Little tricks like this help me stay focused. Technical writing involves enough context switching already -- referring to docs, past articles, research papers, etc. To the extent I can minimize unnecessary context shifting, I do.

All credit goes to https://notbyai.fyi/ for the design of the "not by AI" sticker. I know it is symbolic, but it is personally meaningful to me.

> Allegedly the letter combo TK very rarely occurs otherwise.

Interesting!

When I am writing technical content, I often use placeholders while I'm writing, to stay focused.

I asked around and it turns out a lot of other documentarians do, too! I thought I'd write about this in case anyone finds it helpful!