Not exactly - ChatGPT has two ways it can run Python code. It can use Pyodide and run it directly in the user's browser (for Canvas), and it can also run Python code on one of their servers in a Jupyter environment in a locked-down Kubernetes container (their "Code Interpreter" tool).
To my knowledge they don't yet have a run-Python-in-WASM-on-the-server implementation.
It’s really been advertised heavily lately but I just discovered it a couple weeks ago, and in case you’re unaware the real aha moment with Cursor for me was Composer in Agent mode with Sonnet 3.5.
If you want the highest chance of success, use a reasoning model (o3-mini high, o1 pro, r1, grok 3 thinking mode) to create a detailed outline of how to implement the feature you want, then copy paste that into composer.
It one shots a lot of greenfield stuff.
If you get stuck in a loop on an issue, this prompt I got from twitter tends to work quite well to get you unstuck: "Reflect on 5-7 different possible sources of the problem, distill those down to 1-2 most likely sources, and then add logs to validate your assumptions before we move onto implementing the actual code fix."
Just doing the above gets me through 95% of stuff I try, and then occasionally hopping back out to a reasoning model with the current state of the code, errors, and logs gets me through the last 5%.
On GCP and Azure, most folks would be better off running serverless containers via Cloud Run or Container Apps (AWS has no direct equivalent that scales to 0 and incurs no cost).
Both of these scale to zero and offer 180k vCPU/s free per month, 360k GB/s free per month. You incur billing only against the active execution time. Cloud Run Jobs has a whole separate free monthly grant as well.
You can run A LOT for free within those constraints. Certainly a blog or website. To prevent cold starts, just set up Cloud Scheduler (also free for this purpose) to ping the container every few minutes.
Use Supabase for a DB or one of the serverless options (if it works for your data use case) like Firestore, CosmosDB and you can run workloads for a few cents per month with an architecture that will scale easily if you need it to.
Same. My husky/pyr mix needs a lot of exercise, so I'm outside a minimum of a few hours a day. As a result I do a lot of dictation on my phone.
I put together a script that takes any audio file (mp3, wav), normalizes it, runs it through ggerganov's whisper, and then cleans it up using a local LLM. This has saved me a tremendous amount of time. Even modestly sized 7b parameter models can handle syntactical/grammatical work relatively easily.
If I was in your position, I would ease in to a whole food, plant based diet.
What this means is a lot of cooking from scratch, which means hands! So how about a whole food, plant based diet that requires very little preparation?
This is serious, so please do not waste time with 'cut out teh carbz' bro science. Do not take advice from anyone that talks of 'seed oils' and other keto talking points. Keto and carnivore diets are fad diets that are just another way to get to calorie restriction. They are popular amongst people with protein obsessions and social media influencers, because who does not want to eat steak and butter?
The whole food, plant based diet means no animal products, no refined sugars, no processed foods and lots of plants. Lots is important as vegetables, pulses, grains, beans and fruit are not as calorie dense as a lump of meat. You will need to be eating huge bowls of cooked food and not skipping meals just so you can get your calories in.
On a whole food, plant based diet, you can vary your diet by the season. This means buying from the vegetable and fruit aisles, going for whatever is on offer.
Due to the hands, you might want to buy lots of prepared frozen vegetables. Get the lot.
Oils are what you don't want in your system. Clearly we need some fats but there are plenty in nuts. Personally I only use a small amount of mild olive oil in the air fryer, I don't have butter or fake spreads.
Sugar is surprisingly easy to give up and comes with immediate health benefits as you have to home cook everything to avoid sugar. Sugar is in sauces and other savoury products that you would not expect.
Once you have knocked off sugar, you can knock off the animal products and expand your repertoire of goto plant based recipes.
What works for me is slow cooking. I usually start by putting a chopped onion and some garlic in the pot, to then add some starchy vegetables such as sweet potato, then some leafy greens, then a tonne of lentils and dried beans.
If there is room I put even more vegetables in and add some herbs and spices. Sometimes this could be a curry, or it could be a new herb I am experimenting with. Ginger goes in quite often, there is no fixed recipe as recipes are boring.
I usually add some chopped tomatoes, top up with water and set the thing to do its thing for about four hours.
This approach means I am spending twenty minutes in the kitchen every day, in total. I often add grains such as rice or barley, or I add pasta to the pot after taking my first portion, adding water as appropriate. Grains or pasta does not take four hours, an hour should be good. This means my second portion is a variation on the first.
To top out my slow cooked creation I put some tofu or even some vegetables such as broccoli in the air fryer, with some herbs. This gives different texture.
Just by varying the ingredients I can get variety even though I am doing a one pot meal.
Be an autodidact with this, implement your changes on a monthly basis and see how the inflammation in your hands changes. If you go WFPB then you should end up with excellent gut health, to be in the middle of the Bristol Poo Scale every time, with farts that don't smell.
This is an elimination diet, specifically sugar and animal products. Once you have done the 'factory reset' then you can add in the favourites again, super sensitive to how you feel afterwards. Or you might not want to. I could not care for sugar when it was gone, and the same with dairy, which I thought I was wedded to.
One pot meals, tray bakes and air fried things provide enough variety for me. I don't indulge in salads because of the lack of calories, and neither do I make smoothies because they are for babies, gym bros and people in care homes. Cooking is our original innovation and we need cooked food, mostly starches, to get the calories in.
For anyone interested in starting their own Godot journey, it’s hard to go wrong with Heartbeast. He just posted a new video this week to get a game done _today_!
I'm no good at prompt engineering, so I iterate a lot, rather than trying to come up with the perfect prompt. I start by asking something like,
"please generate a series of short questions and answers that conform to a flashcard format, easy for studying, from these notes"
The responses were too verbose, so I asked:
"Please shorten the answers to short, easily memorizable lines"
That returned pretty good flashcards, so from there I asked it to reformat in a way Anki could consume:
"these are great, please convert these into a code block, where the question and the answer is on the same line, separated by a semicolon and not numbered"
But that put all of the flashcards on the same line, so I added:
"sorry I wasn't clear, each question should be on its own line"
That gave me what I was looking for. The next step is to paste it into Anki, fiddle with the recall and cards-per-day settings, and then get to training!
> The price is a review (even if only a few words) on your preferred social media, either tagged #PricingMoney or with the link jdawiseman.com/PricingMoney.html. There is no paywall, nor registration, nor even cookies, so this price cannot be enforced. Nonetheless, please be fair: please post a review or comment or acknowledgement, tagged or linked or both. Thank you.
Anecdotally, in the very early days Figma rendered its UI entirely in canvas, but we eventually switched over to React for the UI. It ended up being far easier for development cycles and for accessibility/integration with existing browser features.
"Along a less-traveled route, meditation remains what it long was: a deeply transformative pursuit, a devoted metamorphosis of the mind toward increasingly enlightened states."
This pretty much lines up with what meditation has done for me. However, the pursuit of "states" can be a trap in of itself.
As my practice has gotten deeper, I've started to reframe meditation for myself as the process of unrelenting inquiry in the search for base truth. In that pursuit, the practice becomes a process of subtraction of core beliefs and ideas that simply aren't true - or can't be known to be true. As these beliefs disappeared, much of my own personal suffering did as well as so many of the things that were sources of conflict in my mind were predicated on false beliefs.
With this framing meditation can take many forms. Ramana Maharshi famously asked his devotees to start with the question of "Who am I?" and just keep inquiring.
what is a real world use case example of how to finetune, what to feed it during fine tuning, and then how to interact with the model post-fine tuning differently than if you hadn’t fine tuned it?
If you (or anyone else is) interested in the topic, I'd highly recommend giving some of these a look:
Bubeck, Sébastien, Varun Chandrasekaran, Ronen Eldan, Johannes Gehrke, Eric Horvitz, Ece Kamar, Peter Lee, et al. “Sparks of Artificial General Intelligence: Early Experiments with GPT-4.” arXiv, March 27, 2023. http://arxiv.org/abs/2303.12712.
Geoffrey Hinton recently gave a very interesting interview and he specifically wanted to address the "auto-complete" topic: https://youtu.be/qpoRO378qRY?t=1989 Here's another way that Ilya Sutskever recently described it (comparing GPT 4 to 3): https://youtu.be/ZZ0atq2yYJw?t=1656
I'd also recommend this recent Sam Bowman article that does a goood job reviewing some of the surprising recent developments/properties of the current crop of LLMs that's pretty fascinating:
No, I haven’t, though I will eventually. In the meantime, some other pieces that always come to mind as being definitely on that list are Weaponised Sacredness by Sarah Perry, specifically the section on egregores (https://www.ribbonfarm.com/2015/05/07/weaponized-sacredness/ ctrl-f “The Egregore”), Rene Girard’s Scapegoat, and as antagonistic reading (threatening ideas you must engage with) Spandrell’s IQ Shredder, and Nick Land’s Hell-Baked. In general the list is about coming to understand little-recognized forces that control the world.
I used something different. Basic details for your entertainment/edification/comedy source material follow mostly so you can shortcut a comparison if you're building something at your place.
2) Class D amp, Aiyima, Fosi, Loxjie etc Aliexpress is one place to get these. I've used and like Aiyima A03 and their ali store delivers fast.
3) Some nice, high-quality, 2nd hand speakers you like. Wharfedale, JBL, B&W, Acoustic Research, Yamaha. (Or get some active speakers you like and skip #2, eg B&O beolab 6000)
5) configure owntone with your spotify premium, takes less than a minute. (And with your music that you own - takes longer because you take more care).
You now have a multiroom setup with fantastic sound that you can control with http://owntone.local:3689/ including with your phone. And/or you can use the "Retune" app on droid and apple's "itunes remote" app on ios. Better sound than most alternatives for less dollars.
All integrates well with Homeasistant because of course it does.
I really like how mine turned out. Having half a dozen sets of speakers all playing the same music in perfect sync as you move from one room to another while doing chores on the weekend fills me with more joy that I would have guessed. YMMV.
To my knowledge they don't yet have a run-Python-in-WASM-on-the-server implementation.