I’ve also been using
1 Like
There are probably big differences in those different Claudes. The best way to get them to perform is to guide them a bit yourself, by suggesting that “try merge sort instead of bubble sort” or that “do that logic in SQL instead of a selector.”
1 Like
Exactly this. You have to guide and direct a lot and “keep things in check”. And chew problems into a form that makes them easier for AI to solve.
I personally feel that coding is more fun than ever. AI does the \*grunt work and ditch digging, and I can focus on more interesting problems.
3 Likes
I don’t understand why you got angry. Generative LLM models rely on the data used in their training, so they are at their best when solving general, precisely defined problems for which many examples exist, meaning they should indeed be used in a waterfall-like manner. AI is not particularly suited for agile development and creative problem-solving, and the worst is if the context becomes polluted with contradictory prompts, because a person thinks they have explained the matter, when in reality the model’s confusion likely only increases.
We have painstakingly taught people away from the waterfall model, and now we should find a path back to it, because the more text you manage to write before the very first prompt, the better the end result will be. Especially optimization is in its infancy, as the world is full of bad code and far too little material on what well-optimized code looks like and how to achieve it step by step. A solution for this is unlikely to be found with AI tools in the near future either, so projects requiring agility, creativity, and changing specifications will continue to be carried out by skilled coders like you 
6 Likes
Well, the honeymoon with LLMs is coming to an end. OpenAI is rolling out ads to ChatGPT. From there, it will probably expand to all sorts of other data sales to third parties, because they also have to make a living somehow. Sam Altman is my number one “tech villain” in the category. A ruthless guy who will definitely find ways to monetize, no matter how immoral they are. It’s probably a good time to delete all ChatGPT history, if that’s even possible.
17 Likes
Proceeding according to the playbook. The time of free lunches doesn’t last long, and if the product is free, money is collected by monetizing the user base. As you noted, by selling data or with targeted ads.
What a treasure trove query history and intention inference are, for example, for pharmaceutical and supplement companies?
8 Likes
It seems that the one who “wins” this game, measured by market share, is the one willing to incur the biggest losses. Overt advertising drives users away to those who don’t advertise. Data collection is convenient for Alphabet, among others, which can advertise in its other services. But even they probably won’t make money with Gemini.
5 Likes
Alphabet’s profitability is boosted by the fact that Gemini increases the number of solutions built on the Google Cloud stack, thus bringing in high-margin revenue. OpenAI will not be able to compete alone unless it develops a competitor for AWS, Azure, and Google Cloud.
6 Likes
However, seeing my deepest desires in the form of an ad in my private ChatGPT window is certainly not as bad as what social media already is. Social media collects secrets through the camera, microphone, screen, page visits, text written on the phone, location, etc. Somehow this information is then processed to determine what has been found and what can be inferred from it, what has not been found and what can be inferred from the fact that it has not been found. All this information is shared with the user’s relatives and friends through ads and personalized content. For example, if you suffer from hemorrhoids, social media tells your friends about it. It’s another matter whether they grasp the hint in their own news feed, but the information is conveyed. Could this get even worse on an AI platform, since surveillance is already as bad as it can possibly be?
What are the actual threats?
3 Likes
Had to try this. The handwriting changed and the solution doesn’t look right either. I haven’t solved math problems in 25 years, so someone else tell me.
It does pick up the notebook and the task I gave it very well, so in that sense, quite impressive. The task is up there then.
EDIT: let’s try another one.
Prompt: make an image that wins the game. It’s O’s turn and the winning square is very obvious.
The result is like “yeah, right”. The starting situation has also been “slightly” edited.
Edit: let’s add a third gerbil. Original
3 friends. Locations slightly changed. Quite funny, it went over well with the rest of the family.
6 Likes
OpenAI’s story seems to be over? Next, Microsoft will pick up the scraps for a bargain price.
Google was in a similar situation a few years ago, but with an endless war chest, AI culture, etc. OpenAI has nothing. Sam cleared the AI geniuses and culture out of his ego’s way to other companies. Investors are unlikely to want to fund a war against Google.
I’m enjoying this as both a Google shareholder and a Sam Altman hater. Ah, all those arrogant statements, appearing as a non-profit Jesus and an AI visionary.
More information and discussion on Reddit @biomit
https://www.reddit.com/r/stocks/comments/1pck721/openai_declares_code_red_as_google_catches_up_in/
6 Likes
When throwing around comments like these, could you at least provide some background on what it’s about and a summary of the article? There wasn’t even a link.
3 Likes
Below are a couple of excerpts from an FT article regarding Altman’s ‘code red’ announcement.
OpenAI needs every new source of revenue it can find. But the perils of that approach may be starting to show through. Three years after ChatGPT sent Google reeling, the search giant has surged back. Its latest Gemini AI model has been widely lauded, propelling the company’s stock to new heights. Alert to the threat, OpenAI CEO Sam Altman has issued what he called a “code red” to staff, urging them to refocus efforts on ChatGPT.
Google’s ability to use its search engine as an on-ramp to Gemini gives it a powerful customer acquisition tool. Altman’s call for more effort to expand ChatGPT’s capabilities, for instance adding more personalisation and being able to answer a wider range of questions, points to a period of intense competition ahead over new features.
So, Google’s position and strengths are starting to show, with which they have managed to catch up with ChatGPT in Gemini’s user numbers and technology. Conversely, OpenAI has started to branch out into various services in pursuit of additional business, which means that investments and development in its main product, the chatbot, are falling behind the competitor.
5 Likes
Having used Gemini 3 for a while, I’m annoyed by its technical problems:
- “Error: Try again”
- Sometimes it gets stuck in an endless loop: “I see file xyz, I should update it… actually, I see file abc. I see file xyz, I should update it… actually, I see file abc…”
- And Claude fixes linter errors without being asked, whereas Gemini needs to be explicitly commanded, saying, “Hey, there’s a linter error there, fix it.”
2 Likes