A new one comes out artificial intelligence (AI) To the market, social networks and specialized communities hallucinate with their new capabilities, and at the same time, a cycle where users who know the models begin to feel disappointment begins. They begin to see how what until yesterday achieved without problems via Chatbot or API, today stays in a vague attempt. There is a part of sensations, and one real.
“Super broken” models. When he launched, Gemini 2.5 PRO reaped huge praise in social networks. The model was Very fastof the cheapest, had A context window huge and it was A beast in programming. However, for a few weeks, comments have emerged in communities such as Reddit that describe a “unusable” model.
A model, which as described worked incredibly well between March and June, but that after using now at the end of July it released “absolute nonsense“. Showed a conversation with Gemini summarized by the assistant in which he did not stop recognizing errors. Other users also show Examples of behaviors annoying as Do not finish answers.
They are only recent examples of Google’s AI, but even models as praised as Claude have received at different times criticism similareven recently with Claude Code.
Suspicion. Many of the users who have criticized the different models speak of cut models: “My assumption is that they reduced the size of the model,” said A Claude 3.5 user in Hacker News. The suspicion is that, over time, and at times of maximum demand, companies begin to use distilled versions of AI models that are not so intelligent, because they have less dedicated resources to respond to the indications.
Ian Nuttal developer too He observed Claude Code degradationand claimed that he would pay to have a good version that would never be reduced or degraded at peak hours. Alex Finn, also developer, expressed Equally frustration: “This happened to me with all the IA programming tools that I have used.”
It’s not just a sensation. In 2023, many users felt that GPT-4, Openai’s most advanced model at that time, was becoming silly. The company claimed that contrary to what the community denounced, they made each new version “smarter than the previous one.”
However, a Paper Academic It ended with speculation: experts from Berkeley and Stanford checked a spectacular precision drop of GPT-4 among its variants in March and June 2023. In programming, for example “the percentage of generated responses that are directly executable was reduced from 52.0 % in March to 10.0 % in June”. Others statistical studies At the end of 2023 they also showed a significant loss of quality between the December and May model.
Openai and Anthropic confirmed problems. In December 2023, OpenAI recognized that they had received the feedback on the assistant becoming more vague. They claimed that they had not updated the model from a month earlier, and that it was not intentional, recognizing the problem and explaining that “the behavior of the model could be unpredictable.”
Some users came to devise (and achieve, according to their experience) Methods to encourage the model to do betterlike the surprising promise to give a tip or explain to the chatbot that they had no fingers to write the code.
More recently, Anthropic acknowledged To TechCrunch have problems in Claude Code, as slower response times, before complaints of users of having limited use without having affirmed. Users who previously performed tasks normally and now could not progress.
In Xataka | I have tried day, the browser that replaces ARC and bets everything to AI. It hasn’t come out as expected
GIPHY App Key not set. Please check settings