I don't believe you're asking this in good faith because the answer is so obvious. But just in case, it's because OpenAI is by a ridiculously large margin the most well known player in the space and unlike the leaders of other organisations his name is known and he has a personal brand which he puts a lot of effort into promoting.
Why is it obvious. Google is a much more powerful profitable entity that has equally good models. Facebook uses data from Libgen. Both are highly profitable companies who could pay for resources but don't.
Anthropic, the ethical fork of OpenAI doesn't do anything much different nowadays.
OpenAI may have had a head start, but the competition is not far behind.
The two are seen on the picture together! So more similar they could not be. That highlights the irony and hypocrite of capitalism, of better say of human society.
> I wouldn't call o1 a "system". It's a model, but unlike previous models, it's trained to generate a very long chain of thought before returning a final answer
That answer seems to conflict with "in the future we'd like to give users more control over the thinking time".
I've gotten mini to think harder by asking it to, but it didn't make a better answer. Though now I've run out of usage limits for both of them so can't try any more…
not in a way that it is effectively used - in real life all of the papers using CoT compare against a weak baseline and the benefits level off extremely quickly.
nobody except for recent deepmind research has shown test time scaling like o1
I will be happy to use a no-code platforms that lets me eject out into a standard application. Otherwise, I'm just learning and depending on a proprietary system.
Once you need something slightly complex, no-code becomes non-trivial. It requires serious commitment to learn all the techniques the designers came up with.
So most people who use these tools use them for for simple or short-lived apps and side projects, which they could now use web frameworks, but just want to try something new, because they know it's trivial.
Completely agree. No code is great for simple CRUD apps but unfortunately, anything reasonably complex requires code. This is why we ended up building Appsmith which is an open source "low code" application builder. You can write custom JS code across the board & customize your application to build more complex apps. Makes getting started easy but also keeps you going for a long time.
Because a framework with components is just basic UI building blocks. There's a lot more that goes in building a full-fledged app. Layouts, page design, dynamic data, business logic etc. A low code builder like Appsmith helps devs orchestrate all of this much more easily.
I've used WYSIWYG UI dialog editors (mostly the one MSVS used to bundle with C# in... I wanna say 2012?) that maintain their entire state as somewhat-editible two-way code (so if I modified the code, at least within a constrained envelope, it would be reflected in the editor and remain WYSIWYG editable).
Does that count?
If so, I think the lesson is you need to keep the no-code side way, way less ambitious than is theoretically possible in order to keep everyone sane and you might actually have a useful tool that doesn't dig you into a big hole.
Definitely agree with you in terms of limiting functionality built on top of WYSIWYG UIs. 2 years ago I developed a non-trivial app using Retool and it became an unmaintainable spaghetti monster "code" base. (Side note: Retool was pretty solid, my application was just way too complex for it)
The result was I wasted 6 weeks of time when I should've just spent 2-3 days prototyping and then throw out the tool once I had a good idea of what we actually wanted. In the end if didn't matter since the project died anyways :/
PM at Retool, agreed this is hard problem. Like you said, most projects fail so Retool can help you validate and get to success or failure faster. But yeah totally agree that performance, making diffs cleaner, and code search are some things we’re really working on.
Yeah, I had similar experience with https://eclipse.dev/windowbuilder/ in java swing. You can use the drag and drop for most of it, and if you need to hand edit some code, it mostly works and is reflected in the gui tool.
I thought it was pretty productive and honestly sort of miss that kind of development
Mendix does allow this. It runs on JVM containers so from their own "action blocks" you can just say "Call Java action" and execute your code there. There is a java api that you can access from there. Pretty neat and that's how most plugins are built.
On the client side you can do something similar with javascript.
I worked with some platforms that does. However, those usually position themselves as a code generator more than a full application builder.
- Amplication: Last I checked, they allow defining data structure & endpoints etc. before exporting to Nest.js Boilerplate with niceties. They sync to a separate branch on GitHub, so you can use it alongside writing actual logic into the boilerplate, and merge new changes from Amplication to the main branch.
- Draftbit (React Native based Mobile App builder). The builder is not very great, they lack a bunch of features. They make up for this by allowing the developer to add custom JS code very easily. You can write your own Custom React element inside the editor, and with a little massaging things can fall into place. I'm not saying this approach should be abused, though. The team told us once that some of their big clients are mobile dev agencies that mostly just use the platform to Storyboard & bootstrap some UI before doing the actual logic in their own code.
- FlutterFlow (Mobile App builder). I'm biased because I'm still in the progress of helping a client fixing their mess on FlutterFlow. FlutterFlow's code output is really spaghetti. The developer UX on the builder is horrible, so in the end I still opt to eject and deal with the spaghetti code.
If your workflows are exposed to their UI using an API, you can get your workflows out.
Whether it’s worth it vs rebuilding them bespoke is a different question. Or you have to have an ETL layer between your SaaS vendor and whatever open source workflow runner you’d eject to.
Neeva jumped on the AI train after GPT3. The original plan was a privacy-focused Search. People don't pay for that stuff. I can't believe $millions was invested in that idea.
Personalized search - a search engine for all your content based on the data you've collected and visited and private would be something people might pay for.
His books are enjoyable, I'd recommend his most recent one "Four thousand weeks" and the previous one "The Antidote" as well. Especially if you're overwhelmed or gets hit with anxiety often.
