Wrong about Jira

February 5, 2026

Four years ago, I wrote that Jira is good enough. Perhaps not coincidentally, Atlassian stock achieved its all-time high at the time. Four years later, Linear is crushing Jira in the software space. At the same time, Linear doesn’t have significant adoption outside “tech”. I underestimated the exponential curve of software and the imperative of velocity that brings.

It’s a stupidly common mistake. Yes, in a linear world (no pun intended), incumbents like Jira win. But for the last few decades, software has been an exponential force. Linear, which has position itself as the project management tool for software, has effectively bet that software will eat the world faster than Jira’s product can “catch up” to Linear. And so far, that’s been remarkably true. The innovator’s dilemna means that Jira can’t pivot very quickly because it needs to service its legacy, enterprise use cases.

One of Snowflake’s founding stories is a meeting with a big bank (Goldman or something) where they refused to do on-prem. The banker predicted that Snowflake would eventually need to support on-prem, but Snowflake’s founder boldly predicted that the bank would adopt cloud well before Snowflake supported on-prem. Cloud is much more expensive than on-prem, but it is software defined, whereas on-prem is hardware defined or most often, procurement defined (going by the lowest common denominator of corporate velocity). Because of that, Snowflake was eventually proven right: Goldman adopted cloud (and specifically Snowflake) well before Snowflake supported on-prem.

Agentic coding has accelerated software development at least 2x, further pushing software’s advantage over hardware-gated companies: car manufacturers, utilities, etc. With 3D models, you can design a car in software. With World models, you can test self-driving cars in software. With contracted fabs, you can prototype in software.

Underestimating the velocity of change is a typical but obvious mistake. I should know better.