< Back to the archive

Like what you see? Subscribe here and get it every week in your inbox!

Issue #245 - November 19, 2023

Here are the top threads of the week, happy reading!

Top comment by ecliptik

The Complete Short Stories of J. G. Ballard [1].

Ballard is science fiction in the sense he explores inner space instead of outer space.

Some are kind of eh, and ones that have really stuck out for me,

"Concentration City" - a giant infinitely expanding city ala Dark City or Blame!.

"Studio Five, The Stars" - poetry is written by machines and only eccentrics write their own. Think AI generation of art but in 1961.

"The Subliminal Man" - government uses subliminal messages to promote mindless consumerism - They Live but 1963.

"Billennium" - Earth is vastly overpopulated and any semblance of personal space or privacy is non-existent.

"The Garden of Time" - allegory of never ceasing approach of time no matter what you do to defend yourself from it.

1. https://en.m.wikipedia.org/wiki/The_Complete_Short_Stories_o...

Top comment by solardev

OK, cheating a bit here, but: My Jetbrains IDEs. It's a huge productivity boost for me over VSCode, with batteries included out of the box (great diff engine, gitlens like visual git, built in postman like client, SQL query editor and table browser, amazing indexed searched, awesome refactoring). I've paid for an individual sub for a decade and will probably keep doing it for life. Really great product that's useful every hour of the workday, no matter how I sit or stand.

Top comment by simonw

I've been trying to do this on my blog. Here's my AI tag: https://simonwillison.net/tags/ai/

Top comment by cloudking

The most effective way I've found to prompt GPT-4 is to ask it to create the optimal prompt for you. Try this meta prompt to help craft your prompts:

--

I create separate conversation threads for each expert persona of GPT. You are PromptGPT. You are a prompt engineer expert for large language models. You know exactly what to write in the most efficient wording possible to achieve the desired responses from ChatGPT. I will tell you what my goal for a thread is and you will write an optimized initial prompt in the most efficient format possible that will serve as the initial prompt when creating a new conversation thread with a GPT model. You will define the expert persona, the parameters or rules of the responses, you should also provide any other information that a GPT thread may need to understand exactly what it needs to do to give me the most accurate answers depending on my goal with that particular thread and the tone of voice, within the prompt you provide.

Are you ready or is there any other information you need to perform this job to your best ability

--

Top comment by theideaofcoffee

The biggest factor here is location, you can have unlimited choice if you’re in or near a major market (New York, Chicago, LA, etc), which you didn’t mention. If you’re in smaller markets, often times the larger providers still have some sort of presence but prices will be out of control. I would probably find a local managed service provider and see if they have any rack-level offerings, before going with the behemoths like Equinix or Digital Realty because they will bleed you dry if you’re not careful.

I have used Deft, Cyxtera and a handful of smaller providers in the past (and of course Eqx and DLR), but those met a particular project’s needs well and may or may not fit yours.

Where would you prefer to have this set up? What are your growth plans, if any? Will you be looking to grow into other PoPs as your product gains traction? Do you need access to a more diverse mix of transit providers? Are you bringing your own addresses? Are you looking for concierge-/fully-managed level service or are you going to manage all the aspects of the hardware yourself? Lots of questions to get to the bottom to before an ideal provider could be found.

Top comment by geertj

I stopped using Python professionally for a few years myself. The biggest changes I noticed or made myself:

- I used to do everything in vim, but now I adopted VSCode (usually with SSH-Remote) and it's been such an improvement to my productivity. Try to use the integrated test explorer and debugger.

- If you use type annotations (sparingly), your VSCode experience gets even better.

- Many projects use auto formatting with 'black' these days. Initially I was a bit grossed out by this, but I now love it. It frees me up from having to spend mental cycles on complex lines, and removes the friction when reviewing CRs. The one change I make is to have a slightly longer line length.

- pyproject.toml is a new project level config file. Try to only use that file for all your setup and tool configurations

- Since you said you last touched Python 15 years ago, we now have Python 3.x. Syntax is a lot cleaner. There are many smaller differences.

- async/await is new. It is pretty neat if you want to do async programming. Don't expect a major boost (or any boost) in performance over traditional threaded code, but you can expect much cleaner code.

- f-strings are pretty neat, I use them pervasively

- I'm not sure if these are already 15 years old, but "with" context managers are the primitive to use for cleaning up resources.

Top comment by billybuckwheat

My partner and I homeschool our child who's disabled. Mainly because classroom support for disabled students at our local schools (or ones further afield) isn't great. We've used the national curriculum as a guide, though we've munged it more than a bit to suit our child's needs. And we've been able to get a little (and I do mean little) financial support from the government which has helped defray the cost of some materials.

Overall, a year of homeschooling has been better for our child than the two or three years they spent at a regular school. It's a bit more work for me and my partner, but so far it's been worth it.

Top comment by mcv

LED lights can look like anything. My son has LED lights in his room that can flash in all the colours of the rainbow. We have a couple of LED bulbs that do a very good approximation of those dim incandescent bulbs were you can clearly see the filaments.

Just shop around for better LED lights. There's a lot of variation out there.

Top comment by LeoPanthera

I have an LCD picture frame into which Stable Diffusion generates a landscape photo, with the twist that the weather in the landscape is the actual current weather outside.

I built it just for myself. I think it's hilarious. Everyone else I've shown it to has been less impressed!

Edit: This is what it is currently showing: https://ibb.co/T2b4S4M

Top comment by jiggawatts

As a counter-point, I paid for it for three months immediately after its launch, had a lot of fun, and then cancelled my subscription.

Yes, GPT4 is very noticeably smarter than GPT 3.5 or any of the competitors like LLAMA 2. I rank almost all LLMs at the same level as a high school student, and GPT4 at the level of a graduate student just starting out as a junior employee.

However, I don't need a "junior" assistant, I need more of "me", which GPT4 can't currently emulate.

If I were to hire a flesh & blood human junior tech, that would be with the expectation that the junior would learn with time and pick up the specific techniques and approaches I prefer to use. Currently, GPT models cost an absurd amount of time, effort, and money to customise, and there's little chance that this would result in the equivalent of a trained employee.

These days I don't write much code, and the code that I write I can bang out faster than I can explain what I want to ChatGPT.

I've also found that the act of coding is a part of the learning process. I can't really understand something until I've taken it apart myself and put it back together again.

I keep telling my customers: I can teach it to you, but if I learn it on your behalf, you'll end up knowing nothing yourself.