I feel incredibly fortunate to be a part of such a high-agency, smart team, and to work with so many interesting customers on challenging problems & opportunities. Some learnings from a year and a half in at 8090... 1\ There is an immense amount of technical debt in the enterprise and Gov't. These are legacy systems, written in COBOL, Assembly, etc. Everyone knows this, but folks are just starting to come around to the fact that Agents/frontier models can't effectively reverse-engineer and accurately capture requirements for those systems, which means transformation/modernization programs often stall out in the early innings. We've built unique ways to extract accurate business rules at high speed (days/weeks) across many systems. This builds early confidence in the fun part: forward-engineering the new systems, because we're not missing anything the old systems need to do, and dragging timelines out trying to get to ground truth. Ford having to bring back hundreds of engineers with legacy system knowledge is a proof point here... 2\ Time and Materials (T&M) is a fundamentally broken model for legacy modernization. We do our modernization work on a deliverables basis, and deliver outcomes. People want system maps, rules extraction, future architectures and engineering plans, tests & certifications for new systems. They don't want hourly bills for missed milestones and powerpoint slides. 3\ Enterprises have done organizational gymnastics around COTS for too long. They add people, process, and layers to compensate for the functionality gaps from the software they've bought over the years. They want a roadmap of one and are ready to bet on their domain expertise and ideas for how to grow & improve their businesses. We free our customers of the unsexy but important jobs to be done of capturing/writing detailed requirements, building engineering plans, and building/hosting custom software for them. Levie is spot on when he calls out that the idiosyncrasies of business workflows are the hardest part of building custom software. We've made good progress at cracking that (with much more to go). 4\ No one wants to give away their domain expertise to closed-source model providers, and no C-level executives believe the agreements they have with frontier labs will truly protect their knowledge from leaking to these firms and ultimately making its way into models for others to benefit from. The way we build at 8090 keeps IP safe. 5\ We are not cheap (our Custom Software and Modernization work all starts at 7 figures/yr), but customers are willing to make meaningful investments in replacing the key software that powers their business when you show up with a different model, iterate quickly and with quality, listen intently to their needs, and are generally fun to work with. 6\ We are so early. It is very much still Day 1.