GitHub Copilot was originally released in October 2021, four years ago. So much has happened since, it can be challenging to remember what a revelation it was. As has been discussed previously, it wasn’t that the idea itself was without precedent, but the capabilities, the scope and the scale were without peer. Though the concept of a pair programmer that is available 24 hours a day, seven days a week is table stakes in 2025, in 2021 it was, for many developers, mind blowing.
A little over a year after that, however, OpenAI introduced ChatGPT, the interface to its generative model that could produce code like Copilot, but handle a nearly unlimited list of other tasks – if often imperfectly. Thus opened the era of Large Language Models (LLM), the mercurial era we continue to speedrun through today.
The LLM era has seen models embedded in every software and hardware format in existence, and now is increasingly flowing the other way with software being embedded into models. The large frontier models remain almost infinitely versatile, capable of handling almost any conceivable workload from full real-time speech inputs to text-to-video outputs. The narrower domain of coding assistance, however, opened up widely by Copilot years ago, has likewise evolved rapidly. As it’s evolved, it’s upended core assumptions about the development tools market, most obviously that such tools must be free and would command the loyalty of their users.
The promiscuity of developer tool adoption, in fact, has led to short, accelerating cycles in which a new tool with differentiated capabilities emerges, only to shortly be eclipsed by another new tool with new capabilities and approaches. Copilot had its year before ChatGPT (Nov 2022), which had its year before Cursor (Oct 2023), which had its year before Windsurf (Nov 2024). And that’s without getting into the literally dozens of other tools and approaches on the market, a non-exhaustive list of which would include Aboard, Bob, Bolt, Cline, Claude / Code, Gemini / Jules / CLI, Factory, Kiro, Lovable, Poolside, Replit, Same.dev, vibes.diy, v0 and the list goes on.
Each new tool that has its moment in the sun, however, seems to fly a bit too close to it and inevitably, if potentially temporarily, fall back to earth. Sometimes it’s because of the introduction of another breakthrough tool. Sometimes it’s because the team you’re competing against is working their team to death. And sometimes it’s because the bill for tokens comes due. Regardless, it makes for a market about as predictable as the weather in New England.
Which brings us back to this year’s GitHub Universe, though the event was held on the opposite, and much warmer, coast. There was no launch the magnitude of Copilot at Universe this year. Though the company behind the scenes is thinking hard about what the future of development looks like, this year’s announcements – from Agent HQ to Code Quality to Mission Control to Plan Mode – were more about raising the capabilities floor than its ceiling. And for those expecting a new Copilot, this might have been a let down. That simple conclusion can obscure some subtle, more important takeaways from this year’s event, however.
- First, it’s clear that seven years into its acquisition, Microsoft and GitHub are becoming more closely intertwined. Arguably the clearest sign of this was when CEO Thomas Dohmke departed in August and was not replaced, but even at Universe Microsoft personnel were much more visible and tightly integrated into the event than in years past. A greater Microsoft presence does not come without risks, but it also brings immense resources, operational capabilities and – as always – nearly unlimited enterprise account access. In a market that is moving from FOMO to ROI, those things matter. Early signs as well are that as much as Microsoft is integrating with GitHub, GitHub is likewise integrating with Microsoft.
- Second, GitHub is shipping again. Team after team, product after product, GitHub has shifted from a cycle of refinement to one of shipping at velocity. The list of features and enhancements announced at Universe numbered in the hundreds. Some of this is driven by competition with the aforementioned competitors actively unconcerned about burning out their teams, but in many cases it’s simply an executive mandate to ship and ship often. Every software company cycles through periods where they ship and periods where they polish, and in the wake of Universe it’s clear that GitHub is in the former.
- Last, there is the landscape. The years since Copilot debuted have been something akin to an industry fever dream, in which an unending flow of seemingly magical new capabilities let loose vast spigots of capital investment. Developers flitted from tool to tool with Bohemian abandon. Enterprises said get AI and we’ll figure out what to do with it later. Looking around at the market, however, it is now later. Budgets matter suddenly. Buyers are shifting their gazes from potential to measurable impact. Procurement is pushing for vendor consolidation. In such a climate, the combination of GitHub and Microsoft not only represent a wide range of technical capability, but predictability and stability from an enterprise perspective. Frothy, bubbling markets have a tendency to benefit the incumbents, and in the developer tools space no one is more incumbent than GitHub.
GitHub Universe 2025 may not have broken much new ground from a product standpoint, then, but it was nevertheless a crucially important event for understanding where the company and its parent are headed, and as a consequence, where the industry around them is headed. As for Copilot more specifically, you can’t call it a comeback because it’s been here for years, but if GitHub can keep shipping, the code assist landscape will get a lot more interesting, and soon.
Disclosure: AWS (Kiro), GitHub, Google (Gemini et al), IBM (Bob) and Microsoft are RedMonk customers. Aboard, Anthropic, Bolt, Cline, Cursor, Factory, Lovable, OpenAI, Poolside, Replit, Same.dev, Vercel and vibes.diy are not currently customers.