AI agents are science fiction that are not yet ready for thems

This is JumpingA weekly newsletter that breaks one essential story from Tech world. More about all things, Follow Haiden Field. The display arrives in our mailboxes of our subscribers at 8:00. Optimize for Jumping here.

Everything started with Jarvis, that Jarvis the one from Marvelli films.

Well maybe it’s not start With the ARI MAN assistants, but the fictitious system definitely helped the concept of AI agent. Whenever I interviewed AI industry, and Agent AI, they often indicate Jarvis as an example of the ideal and how you need to analyze and find outputs and can offer strategic data on specific aspects of your business. People sometimes do not agree on the correct definition of AI agent, but in its core it is a step beyond chatbot that it is a system that can perform multiple, complex tasks on your behalf, without constantly needing communication with you. It essentially seems to be in your own lists of lubricated masks that need to be met in order to bring to your desired end of the goal. That fantasy is closer to reality in many ways, but when it comes to actual usefulness for everyday users, there are a lot of things that don’t work – and may never work.

The term “a agent” for a long time but it is especially Started in trend in the technical industry in 2023. years. That was a year concept AI agents; The expression was on everyone’s lapping, because people tried to judge an idea and how to do it a reality, but you haven’t seen many successful use cases. Next year, 2024. years, it was a year deployment – People really put the code in the field and saw what it could do. (The answer was … not much at the time. And he filled with a crowd error message.)

I can specify hyper around AI agents that become widespread to one specific announcement: in February 2024, Clarna, Fintech Company, said After one month, his AI assistant (launched Openai technique) successfully conducted 700 full-time services and automated two-thirds of Chert Chats Chats Chats. For months, these statistics appeared in almost every program and the industry I had.

HIPE never died, and in the coming months, every major technical executive director seemed to be in an expression in every earn. Executing Amazon, Meta, Google, Microsoft and the entire hosts of other companies began to talk about their commitment to building useful and successful agian agents – and they tried to put their money where their mouths.

The vision was that one day, and an agent can do everything from booking your trip to create visa for your business presentations. The ideal tool could even say, find a good time and place to hang out with all your calendars, food preferences and dietary restrictions and create a calendar event for everyone.

Now let’s talk about it “AI coding” of everything: There were coding for years carrying Agent and industry. If you asked someone about real life, successful, annoyed use cases for AI agents that don’t happen immediately And not conceptually in an unworthy future, they would point to AI coding – and that was a fairly the only concrete thing that could indicate. Many engineers use AI coding agents and see them as objectively pretty well. Good enough, in fact, on Microsoft and Google, Up to 30 percent The Code is now written by AI agents. And to run like Openai and anthropy, which burns through cash at high prices, one of their largest revenue generators are ai coding tools for clients’ enterprises.

Thus, until recently, and coding was the main case in real life than real life agents, but obviously, it was not washed for everyday consumer. Vision, you remember, was always on AI AI trading agent for “all kinds.” And we’re not right there yet – but in 2025. We approached the years than we’ve ever been earlier.

Last October, Anthropic hit things with the introduction of “Computer usage,” who allowed Claude to use a computer like man, browsing, searching, accessing different platforms and completing complex tasks on behalf of the user. The general consensus was that the tool was a step forward technology, but the examinations said that in practice, he left a lot to want. Fast forward until January 2025. and Openai published an operator, his version of the same thing and charged him as a tool for filling out, grocery order, traveling and creating memories. Once again, in practice, many users agreed that the tool was noisy, slow and not always effective. But again, it was a significant step. The next month, Openai has published a deep research, agent and a tool that could compose long-term reports on any subject for the user, and it has thrown forward. Some said that research reports were more impressive than the content, but others were seriously impressed. And then in July Openai combined deep research and operator into one AI agent product: Chatgpt agent. Was it better than most agent tools facing the consumer who came earlier? Absolutely. Is it still difficult to do the job successfully successfully? Absolutely.

So, there is a long way to reach this vision of the ideal and agent, but at the same time we are technically closer than ever before. Therefore, the technical companies set more and more money in the AI ​​agency, through investment in additional calculation, research and development or talent. Google recently rented a general director of Windsurf, Cofaunder and some members of the research and development team, especially in order to help Google attack their AI agent projects forward. And companies like Anthropic and Openay run each other in ladders, knocking from Rung, to introduce incremental functions to put these agents into the hands of consumers. (Anthropic, for example, just announced Chrome extension for Claude that allows you to work in your search engine.)

So real, what happens next is that we will see and encoding still improves (and unfortunately, potentially replace the job of many) Software engineers at the level). We will also see agents who face the consumer improve, probably slowly but for sure. And we will see agents who are increasingly used for the company and government applications, especially since anthropia, Openai, and Xai have all debit government and platforms in the last months.

All in all, expect to see more false beginnings, starts and stops both mergers and acquisitions such as AI agent competition took (and Hype Mehur continues to deal with). One question we will all have to set up how months we continue: What do we actually want a conceptual “and agent” so you can do for us? Do we want them to replace only logistics or more personal, human aspects of life (ie helping write a wedding toast or a note to deliver flowers)? And how good are they in helping logistics in relation to personal things? (Answer for the last: is not very good at the moment.)