A problem with SaaS is that people use many different types of them, with very different UIs, terminology and workflows. But these inconsistencies are the price to bring about innovation through freedom from restrictions.
I advocate writing a design before you code anything non-trivial. The reason is that it saves you time by clarifying how your code should work.
This can be easy to ignore when you work on your own projects and don't answer to a process.
"Measure twice, cut once."
It looks like Claude 3 has taken the lead as the most intelligent LLM. I saw some arguing that GPT-4 was still better for certain edge cases, but overall the consensus is that Claude 3 is better.
The top LLMs could look completely different by the end of the year.
With Google Gemini 1.5, OpenAI are in danger of losing the LLM crown. I think they'll still have a solid brand, but perhaps not the best at anything. The competition in AI is too ferocious.
But the first mover advantage they had will likely have very long-term benefits.
@spence_freeruns
Good tech docs are a huge plus for any product with a technical aspect. I'm keeping this in mind as I write the docs for AI Construx.
I also think this is an area where user interviews could be huge plus, as sometimes devs can't see their docs through the eyes of a new user.
I'm working on something small for a Google Hackathon (with someone on the idea side).
I wanted to use AI Construx to build it, but it's not quite there yet. I can see quite starkly what's missing to get there though (in theory!). Not that much.
Now that text to video is becoming realistic (), who has ideas for apps that build on it? You don't need to say what the idea is, just wondering how much creativity could be unlocked.
#OpenAI
#Sora
Ever hear of DAOs? Decentralized organizations, a use-case for blockchains. I think I've seen examples of these for game companies recently. Although they don't seem to be 100% decentralized, there's usually a core team somewhere.
How do you know when you're MVP is ready to launch? Perhaps another way of asking, how do you know which features should be included, and which can be delayed until later?
I don't know about anyone else, but I'm not worried about OpenAI forging a path to AGI. They haven't given me any reason to be concerned.
But there are a ton of people that stand to gain from any negative perception about them. Usually in the form of attention.
I'm aiming to launch the AI Construx public beta by April 22nd. This is quite a big project, but it's still quite minimal compared to what it could be.
I'm also working on a smaller Hackathon project (team of 2), which uses Gemini to help salespeople find better leads.
GPT-4o is a welcome incremental improvement from OpenAI.
But it seriously looks like LLM intelligence improvements are greatly slowing and solidifying around a broad set of useful features.
This is a big relief for people worried about their jobs.
I was just reading an article on "drowning in code" ().
I believe that dynamic documentation is the way forward when it comes to this. Auto Wiki by Mutable AI looks very promising.
I've tried Claude 3.5 Sonnet out a bit, and I'm impressed. It definitely seems more intelligent than GPT4o.
Now OpenAI, Google and others really have their work cut out for them. Competition is great. What will the AIs of 5 years from now be like?
My phone just reported a GPT-4 variant, great! Now it's just GPT-3.5 on the desktop (where I am). The roll-out is expanding, which is needed, because GPT-4 based LLMs should be the minimum.
LLMs will become more intelligent and more integrated with other types of AI/ML, leading to AGI.
We'll continue to cooperate with them, becoming far more productive than today.
@agazdecki
Some of us want to validate multiple ideas at the same time, to pursue what has product/market fit. That's what I'm doing, and I'm getting better at it.
Sunset projects nobody wants early. Don't go deep with an unvalidated project.
I feel like I'm working on too many projects, my efforts are too spread out.
I usually overcome this by prioritizing one project, and leave 1/2 days for the others.
Any other tips for a situation like this?
@tewy
I think that a really good all-round health app would be better. Mental and physical health share a lot of practices such as diet.
But you'd need really good marketing to catch anyone's attention. So many apps out there.
@kazuki_sf_
@_Glasp
ChatGPT chimes in... Glasp is a free browser extension-based social web highlighter that allows users to highlight and organize quotes and ideas from the web and share them with others, democratizing access to collective learning.
This week I've been defining the business roles used by CxO on which to base advice. I've tried to simplify the defaults while allowing users to micromanage which roles they are responsible for.
The screens are built in
#Flutter
and look really good!
#BuildInPublic
@kelbels_
Here are some good things to use ChatGPT for when coding:
1. Get feedback on virtually anything you're planning on doing.
2. If you're struggling to get code working, see if it knows a solution.
3. Explaining existing code, or code that I found on a website somewhere.
@chfsrh
Full-stack developer here. Next.js + PostgreSQL is my current stack.
I'm somewhat of a generalist though. I've done consulting, performance tuning, back-end and database development, game dev when I was younger, etc.