What You Write When You Blow the First Half of 2026

The moment I concluded that short-form video was rotting my brain, I deleted YouTube, Instagram, and Facebook off my phone. The next day I caught myself scrolling Instagram in the mobile browser.

Because of that slip, I stumbled onto a post titled “The dance you do when you blow the first half of 2026.” It was oddly comforting. I never assumed building a startup in an unfamiliar country would be easy, but I guess people only really learn how much something hurts once they get hit.

[embed]

Runbear’s H1 2026 Retrospective

When Korean national team crashed out of the World Cup and everyone piled on the head coach, it did not feel like someone else’s story. Whether it is an absence of strategy or a failure of strategy, when the results do not come, you owe yourself a brutal look in the mirror. Exposing your own weak spots is never easy. But if it were easy, everyone would do it. Life is a story anyway. The harder the moment, the more you have to expose it and push through, because the harder the climb, the better the view you get to brag about later.

Runbear’s revenue growth slowed in the first half. Organic growth, the weapon we relied on, cooled off. New signups fell, low-engagement customers churned, and growth stalled. Thankfully, usage from our high-engagement customers spiked sharply and defended our revenue. (For reference, we hit $500K ARR in Q1.)

The problem was clear: it was me.

An organization inevitably follows the direction of its founder. In the first half, a version of me who wanted exponential growth over linear growth shipped a feature under the respectable-sounding banner of “land and expand”: Runbear Personal Agent, a Fyxer for Slack.

The logic was simple. People get too many messages in Slack and suffer for it. If AI used the tools I already have access to, checked information ahead of time, and drafted replies for me, we could efficiently automate human-to-human communication. Fyxer drafts your emails in advance and grew into one of the fastest-revenue startups around. Our core customers do more internal communication in Slack than in email, so if we just made that easier… makes sense, right?

Everything makes sense on paper. Ideas are always the cheapest part and execution is everything. This failure was no exception.

• We did not validate the customer problem first. We did not interview them, check whether they actually wanted this, or verify the job they were really trying to get done.
• We did not focus the team on one thing. We strengthened enterprise support and ran a B2C-to-B2B land-and-expand motion at the same time. On top of that, we ran a vertical-industry strategy in parallel. The messaging got mixed, the onboarding got mixed, and the work got mixed.
• We did not keep doing the things that worked. Publishing use cases, influencer collaborations, the things that were already working, all got neglected under the excuse of “radical change.”
• We outsourced too much. To run paid marketing, outbound email campaigns, cold calling, and more all at once, we leaned heavily on outside help. That is our fault, not the vendors’. We brought in outside hands before we were ready to decide and execute well enough ourselves.

The net result: our AEO broke (SEO survived), and our onboarding funnel broke. Fortunately, our existing high-engagement customers love Runbear more than ever, but that was not because of the Personal Agent.

So what did we learn?

Principles are principles because they matter. The principles we set are a) launch and learn, and b) customer-centric. Ship a beta within two or three days and learn from customers, or if it is a bigger build, go meet the customer first. Budget two weeks of development and it turns into a month to ship and another two or three months to promote.

You need diversity. I am not talking about race, gender, or role. Boris, the creator of Claude Code, recently sorted team members into five types: Prototyper, Builder, Sweeper, Grower, Maintainer. As a Prototyper, I constantly try new ideas, new products, new GTM. Naturally, the existing work that already functions quietly drops down the priority list. The problem is that I am the founder. My priorities seep into the team’s priorities. I need to properly delegate to, and back, the people who steadily keep the working things working, and make them work even better.

Strategy is the combination of “which field will we play on” and “how will we win,” and that is not something you can validate before you execute. If you cannot decide which battlefield to fight on, you try to do everything and pull a muscle. Our team recently paused everything for a week and thought through our battlefield together.

Cut what needs cutting, decisively. This is one of the things I am worst at. I can only let go by relentlessly ordering myself to.

• No B2C. Whether it is PLG or a sales motion, we do B2B properly.
• No general-purpose building tools. Not a tool that can build anything, but a tool that makes a specific job easy to get done.
• Minimize the effort spent on outbound cold email, LinkedIn, and calling. We find B2B customers through inbound, referrals, and conferences.

A note on U.S. verticals. While experimenting with vertical industries, we helped traditional businesses adopt AI, including an event AV rental customer and a customer that sells oatmeal. These businesses usually run on very outdated software, which makes AI adoption hard, and that is a clear opportunity. But it also makes GTM hard. Reaching them through search or ads is extremely difficult. We tried more than 600 cold calls, and even getting to a scheduled meeting took serious effort. The U.S. is promising because each vertical’s market is large enough on its own, but from a GTM standpoint it demands the long, steady grind of conferences, in-person meetings, and warm introductions.

So what about Runbear’s second half of 2026?

It is going to shine.

Claude Tag launched recently. My first reaction was “ugh.” By the next day it had turned into “oh yes.” Getting people to treat AI like a teammate, and to imagine how it works alongside them, is harder than it sounds. Thanks to Claude Tag, the market is now paying attention to exactly that. When the wave comes, you grab your surfboard.

Of course, it is hard for us to go toe-to-toe with Anthropic on the same battlefield. But we have learned more, over a longer stretch, in this field, and we know what the champions inside mid-market B2B companies want and why they need it.

Runbear will focus not on a general per-channel AI agent, but on use-case-driven team agents. What we need is not more features but more customers, and what customers need is to get the job done, not a tool for building AI.

Runbear vs. Claude Tag

We execute fast, we fail fast, and we stack small wins.

P.S. Twenty percent of a founder’s time always goes to finding the next opportunity.