@codepetence@dabeaz Very real problem. People just halt after like 3rd depth of view source. Modern frameworks require you to go till least 10 levels deep in function call stack.
@sadhlife Would love to colab on this. I’ve been experimenting with retrieval augmented gpt3. Could be a great addition to this tool!
https://t.co/J7F9mJhVeC
@bernhardsson Django - because it changes how I think about data and UIs. Both data and UI are now Models (because of django admin)
Streamlit - The execution model is so weird and uncanny, wrapping your head around it is very much akin to learning a new language
@weaviate_io@LangChainAI@ecardenas300 This is cool! Is there a way one can compare the performance of these approaches on a set of example dataset and queries?
@Teknium1 You can ask GPT to rewrite the input query so it's more accessible for classical search engines.
If you use embeddings-based vector search, then using this technique allows you to have a better overlap of the query with the source distribution.
E.g. - https://t.co/2FsLMDzA0G
@codepetence I see you discuss the divide between server vs client side components in nextjs, was wondering if you’ve experimented with @remix_run and does it close that gap well? Other thoughts?
@sama Ability to do custom prompts?
Built a collective of conversational bots on GPT3, but there seems to be no way to give custom scripts to ChatGPT and I would love to migrate!
https://t.co/FeaETFoGM9
@mxcl I have great respect for you so instead of getting into a personal argument, will you please educate me? Seems like I might be missing something big here...
@mxcl CompSci doesn’t become irrelevant unless language models learn mathematics and reasoning which is a hard prerequisite, and something they’re terrible at.
@ramsri_goutham For (1) I've experienced that these services keep the popular models in memory so it's fast. But if you try to deploy your own, the numbers don't match
@ramsri_goutham The 2 biggest questions that are seemingly very hard to find in any FAQs -
1) How do they handle the cold start of models (initial load times for large models can take minutes)
2) What GPUs do I actually get?
@KithRags@WishArya I think bengaluru has a lot of these, but they’re inaccessible if you’re not in the right circles. I recently started curating music events here - https://t.co/fvH2cAiIa6
@silverhairs7@lockwoodw Wrong. The question here is - "trigger single events (navigation/dialogs/etc) as a result of an async operation". In a modern async, procedural language - you "await" for an "async" operation, and then simply write code to trigger "events" as a result of that operation.
@tejpochiraju@LineageAndroid@DunzoIt@swiggy_in Not sure about the code, but the network inspector is magical. Even if the UPI spec is public, the specific endpoints to actually make a transaction are NOT. Plus client code is useful for reverse engineering custom fingerprinting algorithms.
@tejpochiraju@theCuriousMonk@dprank@aksanoble These features are now part of modern dynamic languages like python and ruby. Whats not so clear is how to write reliable code using this magic!
@lzsthw I was recently faced with the option of marketing an open source library so it gets moar stars, recognition, etc.
Instead, I chose to flex the claim of "this library is currently being shipped with 3 apps in production with real users", which paid me handsomely.
@mxcl Hell no, I want my local electrician to be able to fix my fan when it breaks. I don't enjoy replacing the entire unit just because a capacitor reached EOL.
@garybernhardt@exec_prog Had a similar use for a graph structure for automatically saving all related objects in django orm. Surprised that django doesn't do this by itself.
https://t.co/nYGS3JsrJg
@lzsthw AWS probably owes the most amount of money, ever for a company. I can't believe how crappy their client SDKs are.
Case in point - https://t.co/HoDzhwYfQY
@dabeaz "We didn't have money to buy a machine for everyone" -- Imagine telling someone from that time that Google colab will be available for free in 20 years.
I get that macos in the cloud is expensive, but there was no reason for you to put "free" on that button, expect a cheap strategy to get me on your platform. I spent at least an hour learning your random config.yml format hoping I could build macos apps for free.
@progmofo How else does one make money with technology? The more I think about it, the more I notice that most of capitalism (in tech) works because someone is trying hard to hide _something_.
@garybernhardt This is awesome, Gary! As i understand it, the path is defined by a human instructor, using a text file? Do you think there's a way students can choose their own path OR the system can generate path based on specific needs?
@paulg At an admission rate of ~0.7%, this exam fuels a ~200 billion dollar industry - selling a pipe dream to every kid in a country that's already in a very bad shape, academically, economically and even politically.
@desicounty@paulg What about the millions of kids abandoning school to grind in a factory for a min of 2 years, engineered just to pass a standardized test for "top institutes" in India?
@ericsnowcrntly (That article describes how erl interpreter can do preemptively switched multicore green processes using multiple interpreter instances running.)
@ericsnowcrntly I can’t help but notice the similarity with Erlang. I really hope we can have some functionality from that beauty over in python. https://t.co/yDO22yrtuB
@garybernhardt I just wrote a solution to the watcher problem using simple RPC. The actual watcher just listens for a RPC from anywhere. In my case, I had intellij issue a simple `python -c ...` which triggered the compiler in milliseconds. Its amazing how well this works.
@KlimentS@MKBHD The SSD was always a step up in every way. Everyone likes faster drives. But the wireless life earbuds seem to be a PITA to charge up. One more battery to charge!!
@Benjie@garybernhardt > exception handling
There has been numerous times when I wish JS would just tell me what I did wrong, rather than doing "best effort" and coming up with weird answers to simple questions.
@joeerl I'm trying to toy with this idea where you always access the latest version because the object itself is communicating for you in the background. Its not perfect and has a lot of holes. But I feel like it can make things easier for pythonistas.
https://t.co/NBhmgjg6S9
@garybernhardt@jamie_gaskins Maybe this can be done by sending a little itty bitty piece of JS code along with the original webpage? Will take a shot at it, if you haven't found.
@thdxr@garybernhardt We have a bunch of projects doing this in python (especially for data visualisation), the only issue with them is stability and scalability.
@TaranVH NO NO NO, like I meant I can map these buttons and dials to do anything. It's open source, 100 LOC, and cheap -- 5-7$ for that H/W. Compared to the 200$ Pallete gear! And if the SDK is open to everyone, what even is the point of spending 200$
Something went wrong, but don’t fret — let’s give it another shot
You need to unzip your entire archive, not just individual folders, in order to view them here. If you keep experiencing errors, please fill out this form.
Your archive isn’t available here.
You need to view this on a desktop browser in order to see your archive.