Yuval Karmi Is Learning to Get Out of the Way

Yuval Karmi had spent months building his new product. It had features, an audience arriving from Product Hunt, and a premise any busy manager could understand: demonstrate a task once, and let software turn it into a guide. Then people tried to log in. Some could not. Emails were not arriving. Errors were appearing. A product designed to explain how things worked had stumbled at the entrance.

Karmi tells this story himself. That is part of its appeal. The founder offering advice on simple products has supplied the cautionary example, complete with a faulty front door. In his later talk about minimum viable products, he recalled building dozens of features when the product needed far fewer. The essentials had received less attention than the possibilities.

For Karmi, now the Tel Aviv-based founder and CEO of Glitter AI, the episode belongs to a longer education. He has built a venture-backed company, watched it wind down, and started again with a product closely connected to his own mistakes. He is still drawn to the keyboard. He is also learning how much of a company happens away from it.

A company that could explain software

Before Glitter, there was Simpo, which Karmi co-founded with Yohanan Hruschev in 2016. Its customers faced an ordinary problem with expensive consequences: people could not always work out how to use the software in front of them. Product teams needed a way to guide those users without waiting for engineers to build every explanation.

Simpo offered in-product guidance and engagement tools. Its public company profile named customers including Walmart, DuPont, Jet and monday.com. In 2019, it announced a $4.5 million seed round led by Redpoint Ventures, with investors including Seedcamp, UpWest and Elad Gil. There was a recognizable business here, serving recognizable businesses.

Karmi says the company eventually raised $15.5 million and grew to 21 people. Those figures describe a larger undertaking than the initial seed announcement. They also make the ending harder to reduce to the customary line in a founder biography. In 2021, Simpo began winding down. WalkMe purchased its technology on October 4 of that year.

The transaction preserved value in what had been built. Karmi has been explicit that he considers Simpo a failure, and has declined to dress the experience as a success for appearances. He describes the eventual sale of assets and a landing for some employees. That distinction matters to the story he tells next: he returned to building with questions about his own decisions.

The founder in the spreadsheet

One of those decisions was how much work to keep. Karmi could code, write, design and handle operations. Each ability supplied a reason to do another task himself. In his account of delegation at Simpo, the CEO of a 21-person company was spending hours editing Salesforce records and reviewing copy. A useful skill had become a very effective excuse.

The spreadsheet is an unglamorous place for a founder story, which makes it useful. Nobody needs a theory of management to recognize the scene. There is a task. You know how to do it. Explaining it to someone else will take time. You do it yourself, and make the same calculation tomorrow.

Karmi’s verdict on his own role is brief: “I became a bottleneck.” The people he had hired had less room to take responsibility because he kept stepping in. In his retrospective, he acknowledges that work could be done differently from his preferred method and still be done well. Sometimes it could be done better.

The product idea followed a practical sequence. Delegating requires explaining. Explaining requires making knowledge available outside the person who holds it. Karmi wanted to shorten that middle step. His May 2024 launch announcement described software that watches a computer task, listens to the explanation and produces written instructions. “I wish I had this years ago,” he wrote.

“I became a bottleneck.”

Yuval Karmi, reflecting on delegation at Simpo

Giving newcomers a way in

There are other forms of explaining a system to someone entering it. Years before Glitter, Karmi organized a community for people arriving in Israel. After moving back himself, he found no social organization offering the kind of support he wanted for newcomers. In April 2014, he founded Social Tel Aviv.

His account describes free classes, activities and social events. The community reached roughly 2,000 people and continued through events organized by members themselves. It is a useful detail alongside the startup titles: he had spent time creating a setting in which other people could find their footing and eventually organize without him.

He had also volunteered as a web developer for Mishelanu, building its web architecture. His degree, earned at the University of California, Davis between 2008 and 2012, was in Business/Managerial Economics. The coding began much earlier. Karmi says he wrote his first code at about eight years old. Building software had a long head start on running a company.

A smaller company, a familiar temptation

Starting again did not immediately produce Glitter. In 2023, Karmi worked on a project called Conversational Demos. He launched it, then concluded that the sales process would involve the stakeholder meetings and live demonstrations he wanted to avoid. He dropped the project and reused some of its code for the documentation tool.

Glitter’s public launch arrived on May 15, 2024. Karmi’s launch retrospective records a Product Hunt Product of the Day win, early paying customers and questions about integrations. The response gave him evidence that other people recognized the problem. It also supplied plenty of evidence that a real user can discover what a builder has stopped seeing.

The product’s action is straightforward. Someone performs a computer task and explains it aloud. Glitter captures the steps and narration, producing an editable guide with screenshots. When a procedure changes, the guide can change with it. The next person can read at their own pace, with less stopping and starting than a recorded demonstration demands.

The route to launch remained recognizably Karmi’s. More time building. More features. His later MVP lesson asks founders to define the question their simplest product must answer. The advice carries the weight of an admission: he had let the pleasure of making things delay the test of whether anyone wanted them.

Keep the customer, delegate the sorting

Success at attracting attention created another kind of work. In his account of building a feedback system, Karmi recalls an inbox that went from quiet to hundreds of messages. Customers brought bug reports, suggestions and questions. The material was valuable, but it arrived as conversations, scattered across the day.

He connected Intercom to Make.com, used OpenAI to categorize the text and stored the results in Notion. A question about whether the product supported a particular function could become a feature request. Repeated requests could be counted. Each entry retained a route back to the conversation that produced it.

He reports processing more than 2,000 conversations and saving around five hours a week with that system. These are his own operating figures. The more interesting detail is what he chose to remove from his day: the repetitive sorting that sits between hearing a customer and seeing a pattern.

At a Google for Startups Israel and DatA-IL talk, he described a separate support workflow. An agent drafts an answer and sends it to Slack. Karmi reviews it, asks for changes if needed and approves the reply. He wants help preparing the answer while staying involved in the relationship.

The talk opened with a confession. He had tried asking ChatGPT to write the presentation. He disliked the outline. Even an explanation of automation needed someone to decide what was worth saying. Karmi made his own unsuccessful shortcut the opening anecdote, an economical way to establish both his enthusiasm and his limits.

The work he kept putting off

By December 2025, another recurring problem had become difficult to ignore. He kept choosing code over marketing. Features had shipped while the public website lagged behind. Existing customers could discover the improvements; prospective customers had fewer reasons to know they existed. He began investing in feature pages, content, search and a public product updates archive.

His preferred way into marketing is an engineer’s analogy. “The market is just a compiler,” he writes. A failed test can tell him something about a message or an assumption. He can change the input and try again. The comparison gives an uncertain task a familiar procedure.

It also exposes the habit he is trying to interrupt. A developer can always find another worthwhile thing to fix. The codebase offers clear progress and fewer ambiguous conversations. For a founder who enjoys building, another improvement may feel easier than explaining the last one. Karmi’s public writing keeps returning him to the people on the other side of the screen.

An ordinary task, enjoyed

In an October 2025 personal essay, he found something to enjoy in a task few founders would choose for their origin story: writing comparison pages. Putting Glitter beside competitors was engaging, he said. He knew the business might miss his goals. He could still find satisfaction in the work happening that day.

That is a quieter aspiration than the funding totals attached to Simpo. It suits the product he is building now. A task done carefully. An explanation someone else can use. A process made a little easier to repeat. Karmi has given himself another company in which to practice handing work over, and another reason to enjoy the part he keeps.