AINov 9, 2026, 12:13 PM UTC·2 min read

Snipki Founder Shuts Down AI Agent Neo, Warns Against Tool-Hopping

Snipki founder Jens Polomski retired his first AI agent, Neo, after nine months. His takeaway: the idea matters more than the tool, and companies should stop switching.

Founder at a desk reviewing an AI agent dashboard on a laptop, with several AI tool logos in the background. — Image: t3n Magazin

Snipki founder Jens Polomski has switched off Neo, the first AI agent he built, nine months after launching it. The idea behind it lives on, he says, and his main lesson for businesses is to stop jumping between AI tools.

In February, Polomski was among the first to put an agent built on OpenClaw in charge of his email and other work tasks. He assembled a small team of agents, with Neo acting as digital coordinator. Looking back, he argues that no single tool drove the innovation. OpenClaw, in his view, was a combination of existing elements, and the idea behind it carried the progress. Since then, many providers have caught up and built agentic features into their products, spawning several frameworks.

Agents have moved into standard systems

According to Polomski, agents and agentic workflows are now part of standard systems, because major AI providers offer connectivity. His team uses Claude and ChatGPT and also runs its own agents on several servers. The guiding question, he says, is what is useful, not which framework must be used.

Why tool-hopping hurts companies

Polomski advises choosing one tool and sticking with it. He points to the recent rush toward Anthropic, driven largely by its reputation in coding, when many companies switched. Tool-hopping means exporting context from one tool, importing it into another and then back again, only for the picture to change a month later. In a company, he calls that unworkable. Better to decide for specific reasons, such as regional, consumer or usability considerations. How you use a tool matters more and more, he says, than which one you use.

He is also skeptical of aggregator platforms that offer access to many models. In his view the difference lies less in the models than in the harness, meaning the surrounding tool and its agentic workflows, which aggregators tend to add only gradually. Having 20 models to choose from does not necessarily give a company direction.

An orchestrating agent that picks the best model is harder than it sounds, Polomski says: how do you measure "best"? He recalls an auto mode at OpenAI and expects the field to move that way, since efficiency and speed count, and token costs are becoming an issue for more companies. Using AI, he notes, is not free.

Apps worth a look

SUMO: the number puzzle

Tap the right tiles, add them up, hit the target. Easy to learn, hard to master. Free, for all ages.

Dialo: learn languages by speaking

Talk to an AI conversation partner; every sentence comes back corrected and explained in your language. 9 languages, A1 to C2.

ediviewer: read EDI files like documents

Open EDIFACT and X12 messages, see every segment explained, convert to JSON/XML, export PDF and Excel. Runs locally, no upload.

giftdetect: the right gift in minutes

Pick who it is for, the occasion and their personality. The search engine reads product descriptions from local shops and ranks what fits. Free, no sign-up.

Tags:ai agentsopenclawsnipkijens polomskitool hoppinganthropic