# Hope, actually by Irene Sandler January 23, 2026 > _“The real voyage of discovery consists not in seeking new landscapes, but in having new eyes.”_ > _—_ Marcel Proust, _In Search of Lost Time_ It’s the most wonderful time of the year! The whizzing merry-go-round of striving enterprise slows down for a decently-cooked meal, a [holiday movie](https://editorial.rottentomatoes.com/guide/best-christmas-movies/), and some oh-so-precious downtime to reflect and consider the state of affairs in mainframe modernization. What? Not your first choice of subject during the hols? Luckily, I’ve done it for you! Grab a sherry and settle in. ### 1. Mainframe modernization is now _au courant_. It’s the eccentric uncle at turkey dinner whose, er, unusual dress sense is now the buzz of the town. It reflects a grudging acknowledgement that, despite their reliability and sturdiness, mainframes do in many (many) cases represent meaningful constraints on an organization’s ability to grow and adapt. Consider: - Searches for “mainframe modernization” on [Google Trends](https://trends.google.com/trends/explore?date=2022-10-17%202025-11-17&q=mainframe%20modernization&hl=en) have increased markedly since we began in October 2022, even as searches for “mainframe” and “COBOL” (and other similar terms) have stayed flat. - The amount of incumbent vendor activity is through the roof. IBM is doing [everything it can](https://www.allaboutcircuits.com/news/ibm-brings-ai-inside-the-mainframe-with-spyre-ai-accelerator/) to persuade companies that they can “do AI” on mainframes. But as even [Gartner notes](https://www.gartner.com/document-reader/document/7093930), “AI will not solve your mainframe skills struggle.” - Cloud providers are eyeing mainframes, with visions of massive cloud consumption dancing in their heads. Google has a full portfolio of offerings ( [including ours](https://cloud.google.com/blog/products/infrastructure-modernization/accelerate-mainframe-modernization-with-google-cloud-ai)), AWS gave an [early peek](https://aws.amazon.com/transform/mainframe/) at AWS Transform for Mainframes last month, and Microsoft is expanding its Azure Migrate Tool to address mainframes. ### 2. The idea of [modernization as a competency](https://www.mechanical-orchard.com/insights/life-death-and-modernization), rather than a project, is finally taking hold. It’s not yet as fashionable as “mainframe modernization,” but you can see the growing chatter around “continuous modernization.” ### 3. Most new efforts keep circling paths we’ve already explored... and discarded. We’ve noted ( [over](https://www.mechanical-orchard.com/insights/the-need-for-speed) and [over](https://www.mechanical-orchard.com/insights/tale-of-two-complexities) again) that translating code isn’t the cause of modernization failures, it’s the testing and integration uncertainty that happens afterwards. _The problem isn’t a lack of AI, the problem is the conventional codegen-first, fix-later workflow._ *Can AI speed things up? Absolutely, but often with deal-breaking tradeoffs. We’ve used LLMs to help with discovery, only to find their determined non-determinism a roadblock. We’ve tried [pairing AIs](https://www.mechanical-orchard.com/insights/can-two-ais-play-the-tdd-pairing-game) to see if they write better code (not really).* Now AI agents are the new black — black boxes, that is — one each for analysis, documentation, codegen, and so on. We’re experimenting with them, too, but suspect that discrete agents can’t be coupled together in reliably predictable ways, when predictability is precisely what you need. (More on AI agents next year.) --- All that said, we’re seeing another trend: hope. There’s renewed hope in solving this decades-old problem in less painful ways. And while there’s no silver bullet, we’ve been applying our test-driven modernization approach to real workloads, such as: - a manufacturer’s extended warranty system - a bank’s domestic payments system - a healthcare company’s invoice processing system - a retailer’s inventory system Workload by workload, we’re building proof that our approach really is better, safer, and faster: one that delivers on the promise of continuous modernization while minimizing new technical debt. So yes, Virginia, there _is_ a way forward. The real miracle this season is taking something that people fear and loathe and turning it into something routine, even joyful. ## End of year fun stuff If you’ve ever had to take heroic measures to keep things running during the end-of-year craziness, you might enjoy this skit we put together, [Holiday Surge](https://www.youtube.com/watch?v=UFOAjRqr2Vk). We’ve [wrapped up our On a Limb podcast](https://www.mechanical-orchard.com/insights/on-a-limb-2025-wrapped), not just for the year, but for good. It was fun exploring the elements of risk, and how hard it is to orient around preventing risks vs reacting to crises. Next year, we’re planning on diving a bit deeper into the technological aspects of mainframe modernization — it might be a podcast, or it might not be. There’s still time to find that perfect gift! We’re pleased to share our last-minute [Gift Guide for the Panicked](https://www.mechanical-orchard.com/insights/gift-guide-for-the-panicked), especially curated for the people who have had to listen to endless takes on AI and need a break. And last but certainly not least, [Happy Holidays](https://23223845.fs1.hubspotusercontent-na1.net/hubfs/23223845/MO%20HOLIDAYS%202025.mp4?_hsmi=2&utm_content=2) from all of us at Mechanical Orchard!