

Discover more from hrbrmstr's Daily Drop
Drop #236 (2023-04-06): Extended Edition
gh-eco / gh-dash; Extended Prompts; Extensive (List Of) Quarto Extensions
Sure, we acquire or use a tool, app, or utility for what it can directly do for us, but every so often you need just a bit more from it that — by itself — it cannot do. Luckily, allowing third-party components to extend the functionality of apps, tools, utilities, and services has been part of computing almost from the start.
Today, we take a look at three extremely diverse extensions to tools and services you very likely use (or should use!) to crank out projects of all shapes and sizes. One is, admittedly, 100% a stretch for the Drop’s theme, and another actually has two extensions, so I guess it is really three even with the stretch.
gh-eco / gh-dash
GitHub's gh command line utility comes in handy to machinate their API from your terminal (I tend to use hub more, though).
Colby Thomas has made an extension for it — gh-eco — that is, essentially, a terminal browser for GH user landing pages and user activitity. The section header is (obviously) the view of mine.
It's a relatively recent Golang creation, but it has a solid look and feel (thanks to Bubble Tea), and has a bit of extra functionality that renders the READMEs of featured repositories on-demand, in-terminal.
While the current level of functionality is pretty limited, it seems like there's plans to add more components.
On its own, it is also a fine example of how to create an extension for the gh
utility, something we may have to do a walk through in a future Knowledge Drop.
If you like this one, you'll 💙 gh-dash, which I won't spoil by blathering. Just hit the site, do the quick install, and avoid the browser the next time you need to review issues and PRs.
Extended Prompts
It seems crazy that we have (almost overnight) been given1 access to a dizzying array of crazy powerful knowledge work tools.
I am not ashamed to admit that I use ChatGPT (directly and via API) to Get (some) Things Done™, and I suspect many other knowledge workers are doing the same.
I've found just typing a raw direct question into ChatGPT is a fairly useless endeavour. It's another “garbage in, garbage out” situation, and for all but the most trivial of answers, coming up with good prompts — including an extended set of “prompt environment setup” instructions — is super important if you want more reliable responses.
In “Stop doing this on ChatGPT and get ahead of the 99% of its users”, Josep Ferrer emphasizes that the quality of the output depends heavily on the quality of the prompts given to the AI2. He suggests that instead of expecting the AI to think for you, you should be the one doing the thinking and guiding the AI to perform the tasks you need.
Josep provides five main ways to improve the quality of prompts given to the AI. These include:
understanding your needs and requirements
treating AI like a digital intern
creating constraints and avoiding assumptions
adding a specific structure
iterating and refining your input.
The article also provides an example of how to create a good prompt for generating Twitter threads using ChatGPT. The article concludes that to make the most of ChatGPT and similar tools, we must refine our approach and focus on becoming thinkers who guide the AI in executing.
It's also important to realize that the ChatGPT main interface most folks use is pretty horrible. Context-driven AI content generation is more the way to go, and the folks at Human Made recently introduced their work on “WordPress AI”. They call it a “WordPress Copilot” that can “speak” Gutenberg / block-editor.
I use WordPress on my main blog (do not judge), but I despise Gutenberg — their “visual” editor. Markdown ALL THE THINGS, I say! However, their intro video (which I cannot link to directly or embed in Substack) has convinced me to give it a go once they release their prototype. I highly suggest viewing it (it's quick, but it looks super powerful).
Finally, in a recent TC article, Natasha Mascarenhas provides further tips for extending the output capabilities of ChatGPT with tips for better prompt engineering. There's also some gold in a Twitter thread Natasha started on the topic.
Extensive (List Of) Quarto Extensions
I try to keep tabs on additions to Mickaël Canouil's awesome-quarto repo, but I've been declaring RSS bankruptcy more often than not at the end of the day (yay long covid brain fog impairing rapid context switching). Said flushing has made me miss the fact that there are a bonkers number of Quarto extensions out there.
My (work) team's forthcoming microsite and technical blog — launching soon — is based on Quarto, and I need to dig into this list to see what I might be able to use to augment it a bit.
If you don't use Quarto, you are 100% missing out on one of the best publishing tools/ecosystems out there. And, if you — like me — have not been keeping up with awesome-quarto
, you should carve out some time to linger and explore.
FIN
Yesterday's Drop took the place of the traditional WPE version, but I highly recommend taking the weekend off. Y'all deserve a break, and many are likely participating in holiday activities. Trust me, work and projects will 100% be there on Monday. ☮
I mean, ChatGPT does cost actual money, so “given” is a poor choice of word.
Duh
Drop #236 (2023-04-06): Extended Edition
Prompt engineering is seduction.
Thank you for including a section on prompt engineering. Very helpful.