Great list! I also subscribe to Gergeley Orosz' "Pragmatic Engineer" which covers many AI topics now, and to Gary Marcus' substack, which tackles topics more from an LLM skeptic perspective.
Plus I subscribe to updates from Python packages like Langchain and PydanticAI to see what they're up to, since that's usually reflective of what's happening in broader industry.
I'm not on X anymore so I can't follow many of the folks directly, but some of them (like Simon Willison) also post on BlueSky and Mastodon, fortunately. Some folks like Sebastian Raschka and Chip Huyen also post on LinkedIn. Kind of all over, but eventually, I see a good amount of what's happening.
Maybe I’ve been missing some important stuff, but it seems the most relevant and important updates eventually just bubble up to the front page of HN or get mentioned in the comments
The big question for me at the moment is whether to go pay-for-tool or pay-for-model.
If I pay for a tool that includes access to frontier models, then they'll keep the models up to date over time for me, let me use models from multiple providers, and the tool is carefully designed around the capabilities and limitations of the models it works with. On the other hand I can't really use the powerful model the tool works with for other applications or write my own.
If I pay for models, then I can only really use it with that manufacturers tools or tools that aren't optimised for the model but allow you to bring your own keys, and if the model provider I'm paying falls behind then I'm tied in for the duration of the contract. The big advantage is that there is a lot of innovation in tooling happening at the moment and you can avoid being locked out of that or having to pay many times for access to the same frontier models accross multiple different tools.
Andrej's talks have helped me tremendously. They're up on YT [0]. For a long time I used to mentor and help machine learning scientists but when I hear Andrej speak, it's like I'm the student without any knowledge. It was a really strange feeling at first but I've come to value so much. I'm Jon Snow. I know Nothing (compared to Andrej).
Article does not make the case for why you must keep up.
AFAICT the best strategy would have been to completely tune out AI for the last ~3 years:
- AI has not meaningfully improved productivity (unless you’re doing something super basic like react and were already bad at it). If you are using AI in a transformative way, that looks different today than it did 6 months ago.
- AI has not stolen jobs (end of ZIRP did that)
- The field changes so fast that you could completely tune out, and at any moment become up-to-date because the news from 3 months ago is irrelevant.
I don’t get where this meme that “you have to keep up” comes from.
You have agency. You can get off the treadmill. You will be fine.
If you have a decent understanding of how LLMs work (you put in basically every piece of text you can find, get a statistical machine that models text really well, then use contractors to train it to model text in conversational form), then you probably don't need to consume a big diet of ongoing output from PR people, bloggers, thought leaders, and internet rationalists. That seems likely to get you going down some millenarian path that's not helpful.
Despite the feeling that it's a fast-moving field, most of the differences in actual models over the last years are in degree and not kind, and the majority of ongoing work is in tooling and integrations, which you can probably keep up with as it seems useful for your work. Remembering that it's a model of text and is ungrounded goes a long way to discerning what kinds of work it's useful for (where verification of output is either straightforward or unnecessary), and what kinds of work it's not useful for.
To be honest, this even is mostly true in the research side of things. Granted, 99% of research has always been incremental (which is okay! Don't let Reviewer #2 put you off). Lots of papers are filled with fluff. That is, if you have a strong background understanding these systems (honestly, a math background goes a long way to genearlizing this as lots of papers are just "we tried this math idea" and if you already knew it, you'd have a good guess as its effects).
I think it is easy for it to feel like the field is moving fast while it actually isn't. But I learned a lesson where I basically lost a year when I had to take care of my partner. I thought I'd be way behind when coming back but really not much had changed.
I think gaining this perspective can help you "keep up". Even if you are having a hard time now, this might suggest that you just don't have enough depth yet. Which is perfectly okay! Just might encourage you to focus on different things so that you can keep up. You can't stay one step behind if you first don't know how to run. Or insert some other inspirational analogy here. The rush is in your head, not in reality.
Agreed. I played with a few code assistants and I dont see any stark differences in capability. Mostly UI. Do you want it in your editor, on the terminal, in the browser etc. It is because there is fierce competition everything hyped is quite good.
I have a very good idea of how various models work. But the business I run benefits immensely from utilizing the latest models, whether thats ultra low-latency YOLO-style models or “SOTA” high performing ViT, LLMs, etc.
I maintain a funnel sucking up all the PR stuff — but I skip straight to the papers, benchmarks, and githubs.
You still need to learn the names of models, understand their use cases, concepts like MoE, then you have different architectures like diffusion vs transformers, agents etc.
And then you have GenAI like flux and all the open source projects.
I think it's beneficial to get all of that and then keeping an eye on it to catch the moment when it becomes relevant for you and not being surprised and too late.
Just subscribing to OpenAI, Anthropic and Google on YouTube is pretty helpful. They post demos of all major new feature releases that are good to skim through to get a sense of where the frontier is moving (just take all claims about capabilities with a grain of salt).
I've also got some gems from Microsofts Build talks, specifically whenever Scott Hanselman and Mark Russinovich get together, e.g.: https://www.youtube.com/watch?v=KIFDVOXMNDc
Here's my current rule of thumb: If you have successfully built a couple projects using agentic tooling and Claude 4 or similar models: you are doing a fine job of keeping up. Otherwise, you are at least a generation behind.
Don’t you ever let people make you think you have to tirelessly follow some tech progress or stay on top of things all the time.
Progress is like a bus. You can just get on board at any time. You’re not going to fall behind. And staying up to date doesn't keep you “ahead” of anyone.
Doing things is what gets you ahead, and if you don’t feel like doing something right now, don’t worry about, do something later, and you’ll be ahead of people who aren’t doing anything at that moment.
If one wants to follow AI development mostly in the sense of LLMs and associated frontier models, that's an excellent list with over half of the names familiar, to whom I have converged independently.
I have a list in X for AI; it's the best source of information overall on the subject, although some podcasts or RSS feeds directly from the long-form writers would be quite close. (If one is a researcher themselves, then of course it's a must to follow the paper feeds, not commentary or secondary references.)
I'd add https://epoch.ai to the list, on podcasts at least Dwarkesh Patel; on blogs Peter Wildeford (a superforecaster), @omarsar0 aka elvis from DAIR in X, also many researchers directly although some of them like roon or @tszzl are more entertaining than informative.
The point about polluted information environment resonates on me; in general but especially with AI. You get a very incomplete and strange understanding by following something like NYT who seem to concentrate more on politics than technology itself.
Of course there are adjacent areas of ML or AI where the sources would be completely different, say protein or genomics models, or weather models, or research on diffusion, image generation etc. The field is nowadays so large and active that it's hard to grasp everything that is happening on the surface level.
Do you _have_ to follow? Of course not, people over here are just typically curious and willing to follow groundbreaking technological advancements. In some cases like in software development I'd also say just skipping AI is destructive to the career in the long term, although there one can take a tools approach instead of trying to keep track of every announcement. (My work is such that I'm expected to keep track of the whole thing on a general level.)
You don't need to "keep up," you just need to loosely pay attention to identify things/features that will make you more productive, test them out, and keep what actually works (not what some influencer claims to work on social media). In fact, I feel far more confident in my understanding by listening to researchers who dismiss the wild claims about AI's potential—not hype it blindly [1].
There's far too much noise, churn, and indecision at this stage to get any productive value out of riding the bleeding edge.
If it's actually revolutionary, you'll hear about it on HN.
I think AI will have a huge impact but I don’t see a need to keep up. The current systems are still primitive and I bet in a few years they will be way easier to use and more productive. You can use current tech like phones or cars just fine without having kept up. I don’t see why AI will be different.
In a sense it may be more efficient to ignore the current state for a while and jump on later.
I think the feeling to keep up is simply fear of being left behind. Fear is the same thing driving people to become defensive when others dismiss the idea of needing to keep up, because that undermines their core belief that the new “AI skills” they’ve acquired will keep them safe from job disruption.
It’s also interesting that at the heart of the skill set evolving around efficient LLM use, is communication. Engineers and technical people have been the first to admit for decades that they struggle with effective communication. Now everyone’s an expert orator, capable of describing in phenomenal detail to the LLM what they need from it?
Not worrying about all this feels so much better!
This week I saw a video of a robot assembling PCBs at lightning speed. This reminded me a lot of an LLM coding for us, but there are still numerous people in the production line to design, oversee and manage this kind of production. Software engineering is changing but not going.
If you want to stay efficiently up to date with the AI era, I recommend spending five minutes a day on https://infobuzz.ai. The site manually curates AI news from major media outlets within the past 24 hours—super useful.
44 comments
[ 2.5 ms ] story [ 66.7 ms ] threadhttps://newsletter.pragmaticengineer.com/ https://substack.com/@garymarcus
Plus I subscribe to updates from Python packages like Langchain and PydanticAI to see what they're up to, since that's usually reflective of what's happening in broader industry.
I'm not on X anymore so I can't follow many of the folks directly, but some of them (like Simon Willison) also post on BlueSky and Mastodon, fortunately. Some folks like Sebastian Raschka and Chip Huyen also post on LinkedIn. Kind of all over, but eventually, I see a good amount of what's happening.
If I pay for a tool that includes access to frontier models, then they'll keep the models up to date over time for me, let me use models from multiple providers, and the tool is carefully designed around the capabilities and limitations of the models it works with. On the other hand I can't really use the powerful model the tool works with for other applications or write my own.
If I pay for models, then I can only really use it with that manufacturers tools or tools that aren't optimised for the model but allow you to bring your own keys, and if the model provider I'm paying falls behind then I'm tied in for the duration of the contract. The big advantage is that there is a lot of innovation in tooling happening at the moment and you can avoid being locked out of that or having to pay many times for access to the same frontier models accross multiple different tools.
> intended to be a ten minute read that catches you up on the most important developments from the past month in LLMs
https://simonwillison.net/about/
[0] https://www.youtube.com/@AndrejKarpathy
AFAICT the best strategy would have been to completely tune out AI for the last ~3 years:
- AI has not meaningfully improved productivity (unless you’re doing something super basic like react and were already bad at it). If you are using AI in a transformative way, that looks different today than it did 6 months ago. - AI has not stolen jobs (end of ZIRP did that) - The field changes so fast that you could completely tune out, and at any moment become up-to-date because the news from 3 months ago is irrelevant.
I don’t get where this meme that “you have to keep up” comes from.
You have agency. You can get off the treadmill. You will be fine.
Despite the feeling that it's a fast-moving field, most of the differences in actual models over the last years are in degree and not kind, and the majority of ongoing work is in tooling and integrations, which you can probably keep up with as it seems useful for your work. Remembering that it's a model of text and is ungrounded goes a long way to discerning what kinds of work it's useful for (where verification of output is either straightforward or unnecessary), and what kinds of work it's not useful for.
I think it is easy for it to feel like the field is moving fast while it actually isn't. But I learned a lesson where I basically lost a year when I had to take care of my partner. I thought I'd be way behind when coming back but really not much had changed.
I think gaining this perspective can help you "keep up". Even if you are having a hard time now, this might suggest that you just don't have enough depth yet. Which is perfectly okay! Just might encourage you to focus on different things so that you can keep up. You can't stay one step behind if you first don't know how to run. Or insert some other inspirational analogy here. The rush is in your head, not in reality.
I maintain a funnel sucking up all the PR stuff — but I skip straight to the papers, benchmarks, and githubs.
And then you have GenAI like flux and all the open source projects.
I think it's beneficial to get all of that and then keeping an eye on it to catch the moment when it becomes relevant for you and not being surprised and too late.
1. Stop living other people's experiences. Start having your own.
2. Stop reading blogs. Start building apps.
3. Everyone's experience depends on their use case or limitations. Don't follow someone's opinion or ideology without understanding why.
4. Don't waste time chasing employees or researchers on Twitter or Substack. Most of them are just promoting themselves or their company.
5. Don't let anxiety or FOMO take over your time. Focus on learning by doing. If something important comes out, you'll find out eventually.
6. Being informed matters, but being obsessed with information doesn't. Be smart about how you manage your time.
That's what I tell them.
It's fine to go do other things with your precious time instead.
I've also got some gems from Microsofts Build talks, specifically whenever Scott Hanselman and Mark Russinovich get together, e.g.: https://www.youtube.com/watch?v=KIFDVOXMNDc
https://x.com/rowancheung - Rowan Cheung: “Daily” updates and insider access
No, I don't think I do. Been working great for me so far.
Progress is like a bus. You can just get on board at any time. You’re not going to fall behind. And staying up to date doesn't keep you “ahead” of anyone.
Doing things is what gets you ahead, and if you don’t feel like doing something right now, don’t worry about, do something later, and you’ll be ahead of people who aren’t doing anything at that moment.
I have a list in X for AI; it's the best source of information overall on the subject, although some podcasts or RSS feeds directly from the long-form writers would be quite close. (If one is a researcher themselves, then of course it's a must to follow the paper feeds, not commentary or secondary references.)
I'd add https://epoch.ai to the list, on podcasts at least Dwarkesh Patel; on blogs Peter Wildeford (a superforecaster), @omarsar0 aka elvis from DAIR in X, also many researchers directly although some of them like roon or @tszzl are more entertaining than informative.
The point about polluted information environment resonates on me; in general but especially with AI. You get a very incomplete and strange understanding by following something like NYT who seem to concentrate more on politics than technology itself.
Of course there are adjacent areas of ML or AI where the sources would be completely different, say protein or genomics models, or weather models, or research on diffusion, image generation etc. The field is nowadays so large and active that it's hard to grasp everything that is happening on the surface level.
Do you _have_ to follow? Of course not, people over here are just typically curious and willing to follow groundbreaking technological advancements. In some cases like in software development I'd also say just skipping AI is destructive to the career in the long term, although there one can take a tools approach instead of trying to keep track of every announcement. (My work is such that I'm expected to keep track of the whole thing on a general level.)
There's far too much noise, churn, and indecision at this stage to get any productive value out of riding the bleeding edge.
If it's actually revolutionary, you'll hear about it on HN.
[1] https://x.com/burkov
In a sense it may be more efficient to ignore the current state for a while and jump on later.
It’s also interesting that at the heart of the skill set evolving around efficient LLM use, is communication. Engineers and technical people have been the first to admit for decades that they struggle with effective communication. Now everyone’s an expert orator, capable of describing in phenomenal detail to the LLM what they need from it?
Not worrying about all this feels so much better!
This week I saw a video of a robot assembling PCBs at lightning speed. This reminded me a lot of an LLM coding for us, but there are still numerous people in the production line to design, oversee and manage this kind of production. Software engineering is changing but not going.