As one of many stalwarts of the fashionable developer tooling trade, GitLab wants little introduction. The corporate helped popularize the thought of a single platform for managing the software program improvement lifecycle, spanning supply management, CI/CD, safety scanning, collaboration, and deployment inside a single system.
However now, GitLab is restructuring itself for an entire new paradigm: that AI brokers will improve the quantity of software program being constructed — and builders will spend their time supervising, reviewing, and coordinating machine-generated code fairly than writing each line themselves.
Bill Staples, who has served as GitLab’s CEO since co-founder Sid Sijbrandij stepped down in 2024, introduced a slew of adjustments on the firm this week, together with layoffs, govt adjustments, product consolidation, and a renewed give attention to AI-driven software program improvement.
The overhaul comes throughout a troublesome stretch for GitLab. The corporate’s market capitalization has dropped roughly 66% to round $3.7 billion over the previous 15 months, as buyers query how AI will reshape software program improvement and developer tooling extra broadly.
Staples, for his half, thinks he has a solution to quell this decline. In an open letter published on Monday, Staples argues that AI gained’t shrink the software program trade — it’ll broaden it. Making software program cheaper to supply, the pondering goes, will solely create demand for extra of it. Economists name that Jevons’ paradox — a nineteenth-century financial idea, named for a similar effect that saw extra environment friendly steam engines drive coal consumption up, not down — and GitLab is betting its future on it.
“As the price of producing software program collapses, demand for it’ll broaden.”
“Software program has been the drive multiplier behind almost each enterprise transformation of the final 20 years,” Staples writes. “The constraint was the price and time of manufacturing and managing it. That constraint is collapsing. As the price of producing software program collapses, demand for it’ll broaden.”
A “machine-scale” rebuild for the agentic period
Digging into the main points of the letter, Staples factors to a number of massive architectural and organizational bets constructed round what GitLab calls the “agentic period” of software program improvement.
“Software program shall be constructed by machines, directed by individuals,” Staples writes. “Brokers will plan, code, evaluation, deploy, and restore.”
Crucially, Staples argues that this gained’t scale back the significance of engineers, although it does change the place their worth exists. Builders will give attention to higher-level system design, structure, governance, reasoning via failures, and coordinating fleets of AI techniques working throughout the software program lifecycle.
That framing additionally helps clarify why GitLab is placing such heavy emphasis on orchestration. In Staples’ telling, the brand new problem is coordinating massive numbers of brokers working concurrently throughout repositories, pipelines, approvals, deployments, and enterprise coverage techniques.
“Enterprises don’t want agent exercise. They want operating software program that strikes the enterprise ahead. Orchestration is the layer that will get you there.”
“Enterprises don’t want agent exercise,” Staples notes. “They want operating software program that strikes the enterprise ahead. Orchestration is the layer that will get you there.”
To help that shift, GitLab is rebuilding massive elements of its underlying platform for what it describes as “machine-scale” necessities. Staples argues that current developer infrastructure was largely designed round human-paced workflows — particular person builders opening pull requests, triggering pipelines, and committing code at comparatively predictable charges. AI brokers change that dynamic fully.
“Brokers open merge requests in parallel, set off pipelines across the clock, and push commits at a price no human staff ever did,” Staples writes.
GitLab has already been laying the inspiration for this new course, together with its Duo Agent Platform, which launched in January. Talking to The New Stack in February, Staples argued that coding was by no means actually the bottleneck — builders spend solely 10 to twenty% of their day writing code, with the remainder consumed by critiques, pipeline runs, safety scans, and compliance checks. “That code being generated even quicker simply will get caught within the queues that comply with on the coding,” he mentioned. The Duo Agent Platform is GitLab’s try to automate throughout that total lifecycle, not simply the coding half.”
Leaning on its legacy
Among the many firm’s greater bets are rebuilding GitLab into extra API-first, composable providers, creating agent-specific APIs, and redesigning orchestration techniques able to coordinating autonomous software program brokers throughout the complete improvement lifecycle.
However GitLab additionally believes its largest benefit within the AI period might come from one thing older and vaster: the quantity of enterprise context already flowing via its platform.
“Each dev software vendor is converging on comparable code era capabilities. What doesn’t commoditize is the distinctive context the mannequin will get to work with.”
“Each dev software vendor is converging on comparable code era capabilities,” Staples writes. “What doesn’t commoditize is the distinctive context the mannequin will get to work with: an information mannequin that connects planning, code, evaluation, safety, deployment, and operations throughout each challenge and repository, collected over years of a staff’s work.”
The argument is that whereas code era fashions are converging throughout the trade, the organizational context is tougher to breed. The corporate is successfully betting that brokers working inside GitLab’s current ecosystem will make higher choices as a result of they will draw from years of collected buyer workflow knowledge spanning repositories, CI/CD pipelines, deployments, approvals, and operational historical past.
That additionally helps clarify why governance stays such a big a part of the corporate’s positioning towards enterprise clients. Staples presents governance because the mechanism that enables firms to securely deploy bigger numbers of autonomous brokers inside manufacturing techniques.
“Like a race automotive, it doesn’t matter how briskly you possibly can go if you happen to can’t preserve management,” he writes.
The timing, although, is probably considerably awkward for GitLab. The corporate initially emerged as certainly one of GitHub’s most credible rivals earlier than repositioning itself across the broader software program lifecycle and enterprise DevOps tooling. But at the same time as frustration with GitHub has bubbled up throughout elements of the developer group in recent months — starting from reliability complaints to criticism of the platform’s course underneath Microsoft — GitLab hasn’t emerged as a significant beneficiary.
One possible cause is switching prices: groups deeply embedded in GitHub’s CI/CD workflows, integrations, and tooling don’t migrate simply, even when they’re pissed off. Satirically, that very same dynamic — platform stickiness as a aggressive moat — is exactly what GitLab is now betting on with its personal enterprise clients within the AI period. As one group member on X put it: “It’s loopy GitLab isn’t getting greater throughout this GitHub fiasco period.”
“It’s loopy GitLab isn’t getting greater throughout this GitHub fiasco period.”
Jevons’ paradox, a recurring theme
A lot of the normal developer tooling market has traditionally trusted charging per developer seat. AI brokers complicate that equation: they will do the work of many builders, however they don’t want seats.
If software program might be produced at far better quantity and pace, it raises questions on how conventional developer tooling companies seize worth in that world. GitLab’s reply is that AI will improve general software program demand quicker than it reduces the necessity for engineers — and that platforms able to coordinating brokers at machine pace shall be price extra, not much less, consequently.
It’s a familiar argument in AI circles, and one which has discovered vocal help amongst executives resembling Field CEO Aaron Levie, who has made a similar case that software program demand will broaden as manufacturing prices fall.
Blogger and open-source developer Simon Willison writes that “Jevons-paradox-inspired hope” for AI is essentially constant along with his personal pondering, although he cautions that GitLab’s place can be formed by its enterprise incentives — notably at a second when buyers seem unsure about how AI brokers will have an effect on the long-term economics of developer tooling firms.
“In case your total enterprise is determined by software program engineering rising as a area and producing bigger volumes of extra profitable seats, you could have a robust incentive to consider that brokers could have that impact,” Willison writes.
For GitLab, the guess is that Jevons’ idea holds true for the agentic AI period — the choice is just too bleak to contemplate.
YOUTUBE.COM/THENEWSTACK
Tech moves fast, don’t miss an episode. Subscribe to our YouTube
channel to stream all our podcasts, interviews, demos, and more.
