Artificial Intelligence – Stacking Trades

Vibe Coding at a $9 Billion Valuation: The Bet That AI Will Replace the Developer Hiring Cycle

Stacking Trades — Tue, 14 Apr 2026 18:30:01 +0000

In March 2026, Replit raised $400 million at $9 billion — triple what the company was worth six months earlier. Its annual recurring revenue at the time of the raise was $240 million. That is a roughly 37x revenue multiple, priced into a company that did not exist in its current form two years ago and whose core product — letting anyone build a functioning app by describing what they want in plain English — was not commercially viable three years ago.

The number that made investors move was not the valuation. It was the trajectory. Replit’s ARR stood at $2.8 million at the end of 2024. By September 2025 it had reached $150 million annualized. By early 2026, the company was at $240 million and targeting $1 billion by year’s end. That kind of growth curve does not happen in normal enterprise software markets. It happens when a category is being invented in real time, and the companies inside it are capturing demand that had no prior outlet.

The category is vibe coding. And what it is doing to the market for software development — and to the companies that built the last generation of tools — is not a minor disruption. It is a structural repricing of who gets to build software, at what cost, and from whom.

What Vibe Coding Actually Is

The term was coined by Andrej Karpathy, the former OpenAI and Tesla AI lead, to describe a workflow where the developer does not write code so much as direct it — describing what they want, accepting what the AI produces, and iterating through prompts rather than syntax. Karpathy used it to describe his own experience as a seasoned engineer taking a more relaxed approach with AI assistance. The companies building the vibe coding market have taken the concept considerably further.

Replit’s Agent 4, announced alongside the March fundraise, does not present the user with a code editor at all. It replaces the traditional development environment with an interactive canvas — closer to Figma than to VS Code — where users sketch what they want and multiple AI agents execute tasks in parallel: one handling the database, another building the frontend, a third managing authentication. The company claims Agent 4 runs ten times faster than its predecessor, and notably, Replit built Agent 4 using Agent 3. The system is recursive in a way that has no equivalent in conventional software development.

Replit is not alone. Cursor, built by Anysphere, has crossed approximately $2 billion in ARR after raising at a $29.3 billion valuation — the largest in the category. Lovable, a Swedish startup, reached $400 million ARR with over 200,000 new projects created on its platform daily, and announced in March that it is actively pursuing acquisitions to consolidate the market. Cognition’s Devin product, which takes a more autonomous agentic approach to end-to-end coding tasks, raised at a $9 billion valuation after acquiring Windsurf. Vercel, best known for Next.js and its cloud hosting stack, raised $300 million at $9.3 billion on the thesis that its v0 agent — which deploys directly into its existing infrastructure — has an operational moat pure AI coding tools cannot replicate.

The category has attracted more than $5 billion in venture capital since 2024, with the pace accelerating sharply into 2026.

The February Selloff Was the Market Asking a Question

In early February 2026, roughly $285 billion in enterprise software market capitalization disappeared in a matter of weeks in what analysts labeled the SaaSpocalypse. The proximate cause was a cluster of agentic AI announcements, but the underlying logic was vibe coding: if a non-developer can build a functional CRM, project management tool, or internal workflow app from a text prompt in under an hour, the case for paying $50 to $200 per seat per month for a rigid SaaS product becomes harder to make.

The categories hit hardest were horizontal SaaS tools — the ones whose value has always been in packaging functionality, not in deep domain expertise. Vertical software with regulatory complexity, compliance requirements, or specialized data moats was less affected. Healthcare platforms, financial services infrastructure, and government-specific systems held their valuations because their value is not in the UI layer that vibe coding can now generate on demand.

For sophisticated investors, the February selloff was not a verdict. It was a question being priced in real time: which software companies have moats that survive prompt-based app generation, and which ones are selling something that a $15 session can approximate? That question does not have a clean answer yet. But the fact that the market is asking it — loudly, with $285 billion in market cap at stake — tells you something about where institutional money thinks the risk is concentrated. The per-seat model under pressure is playing out on the same fault line.

The Revenue Multiples Require a Specific Bet

Replit at 37x revenue, Cursor at an implied multiple well above 10x on $2 billion ARR — these numbers only make sense if you believe a few things simultaneously. First, that the total addressable market for software creation is about to expand dramatically, not merely shift. Second, that the leading platforms will capture durable market share rather than getting commoditized as the underlying models improve and the cost of inference falls. Third, that enterprise adoption — where gross margins are far healthier than consumer — scales fast enough to justify the valuations before the next funding cycle.

Replit’s own unit economics illustrate the challenge. The company reported gross margins around 23% in mid-2025 across its full user base, well below software industry norms — a direct consequence of the compute costs embedded in running AI agents at scale for 50 million users. Enterprise margins, by Masad’s own account, run closer to 80%. The strategic implication is clear: consumer user counts are the acquisition story, but enterprise contracts are the business. The company’s disclosure that employees at 85% of the Fortune 500 are building on Replit is doing a lot of work in the pitch deck, but it is not the same as saying 85% of the Fortune 500 has an enterprise contract.

Cursor’s position is structurally different. With $2 billion in ARR at a $20 monthly Pro subscription price point, the company has demonstrated it can convert developer adoption into recurring revenue at scale. Its challenge is the reverse of Replit’s: Cursor serves developers who can read the code it produces, which makes it harder to expand the market to non-technical users. OpenAI’s Codex serves over 2 million weekly users, a direct competitor operating from a subsidized cost structure that Cursor cannot match without building its own models — which it is reportedly doing.

What This Means for the Developer Hiring Market

The labor market argument embedded in vibe coding valuations is more consequential than the software market argument, and it is receiving less attention. Klarna’s CEO prototypes ideas himself rather than tasking engineers — filtering his own ideas before they ever reach his technical team. Google CEO Sundar Pichai said publicly that he has been using Replit and Cursor to build personal tools. Replit’s Masad regularly cites the case of a user who built a working ERP for $400 instead of paying a vendor’s quoted price of $150,000.

These are anecdotes. But they point to a structural shift in who initiates software projects, who approves them, and what the minimum viable internal tool looks like. If a product manager can prototype and ship an internal dashboard without an engineering ticket, the demand signal that feeds junior developer hiring weakens at the margin. If a small business can build a customer portal without a contract, a category of development agency work disappears. The platforms are not replacing senior engineers solving genuinely hard problems. They are compressing the long tail of routine software requests that previously required human time to execute.

Gartner projected that 60% of new code will be AI-generated by the end of 2026. Stack Overflow data put the share of developers using AI coding tools daily at 92% as of early 2026. These numbers suggest the baseline has already shifted. The vibe coding platforms are competing not just with each other but with GitHub Copilot, with Claude Code, with every AI coding assistant that hyperscalers and frontier labs are bundling into their existing developer relationships.

“Replit is kind of replacing a lot of the no-code, low-code tools, which really never worked very well. They get initial productivity boosts, but a lot of times that ended up actually slowing down a lot of companies.”<
— Amjad Masad, CEO, Replit, interview with B-17, October 2025

The Competitive Risk Nobody Is Pricing

The vibe coding platforms have a problem that their valuations do not fully reflect: they run on models they do not control. Replit, Cursor, Lovable, and their peers are inference wrappers around Anthropic, OpenAI, Google, and xAI models. When those models improve, the platforms improve — but so does every competitor using the same underlying intelligence. When OpenAI bundles Codex into ChatGPT for $20 a month, or when Anthropic ships Claude Code with capabilities that rival standalone IDEs, the differentiation argument for dedicated vibe coding platforms becomes harder to sustain.

The platforms’ response is to build up-stack and down-stack simultaneously. Replit’s full-stack deployment model — where the app is built, hosted, and monetized within the same environment — creates lock-in that a raw model API cannot replicate. Cursor is building in-house models. Vercel’s deployment infrastructure is the moat. Cognition acquired Windsurf to expand its enterprise footprint. These are real competitive responses, but they are expensive, and they require each platform to win a land grab before the model providers close the gap.

The $9 billion question — repeated across Replit, Cursor, Cognition, and Vercel simultaneously — is whether these platforms have enough of a lead, and enough of a moat, to sustain their valuations when the next round of model releases arrives. The revenue growth says yes. The margin structure and competitive exposure say the jury is still very much out.

What to Watch Next

Replit’s Agent 4 launch, delayed to May 2026 — the first major product test of whether a fully canvas-based, multi-agent development environment converts at scale beyond the early adopter base.
Cursor’s in-house model development timeline. If Anysphere ships a proprietary model competitive with Anthropic and OpenAI, its margin structure changes materially and the $29.3 billion valuation becomes easier to defend.
Enterprise contract disclosures. Both Replit and Lovable have cited Fortune 500 presence; watch for any revenue breakdowns that clarify what share of total ARR comes from enterprise vs. consumer.
Competitive moves from hyperscalers. AWS, Google Cloud, and Microsoft Azure each have developer distribution that none of the vibe coding platforms can match — monitor whether any bundle a comparable product into existing cloud agreements at materially lower price points.
Junior developer hiring data in tech. If vibe coding is compressing demand for routine software work, the signal will show up in job postings and entry-level engineering salary trends before it shows up in any platform’s ARR report.

Big Tech’s AI Spending Is So Big It’s Splitting the Market in Two

Stacking Trades — Tue, 10 Feb 2026 20:13:38 +0000

The numbers coming out of Big Tech this earnings cycle are large enough to change the tone of the entire market conversation. Alphabet has told investors it expects capital expenditures of roughly $175 billion to $185 billion in 2026, nearly double its 2025 level, as it tries to stay ahead of demand for servers, data centers, and networking equipment.

Meta, meanwhile, said it anticipates 2026 capital expenditures in the range of $115 billion to $135 billion, framing the step-up as investment in AI infrastructure and “Meta Superintelligence Labs” efforts.

This is where the market splits. One camp looks at these figures and sees a land grab that will cement long-term winners. The other sees a bill that is arriving faster than the proof.

The spending number that changes the story

When a company spends aggressively, investors usually ask two questions: what is it buying, and when does it pay back. With AI, the first answer is increasingly clear. It is compute capacity, power-hungry data center buildouts, and the specialized hardware and networking that make models usable at scale. Alphabet’s own explanation has been blunt: it has been capacity constrained, and the spending is intended to relieve that.

The second answer is where the argument starts. The payoff can show up as higher cloud revenue, better ad products, faster product cycles, or enterprise demand that sticks. But “can” is not the same as “will,” and the timelines matter when the checks are this large.

Why debt markets suddenly have a front-row seat

There is another reason this feels different than prior tech capex waves. Companies are increasingly financing the buildout in ways that make it visible beyond the equity market. Alphabet’s bond activity, including unusually long-dated debt, has been framed as part of funding its AI expansion.

That matters because it turns AI infrastructure into something the credit market prices, too. When spending is financed, the debate widens from product optimism to balance sheet strategy, cost of capital, and how durable demand needs to be to justify the scale.

The human version of the bull case and the bear case

The bullish view is simple: AI is becoming a general-purpose capability, and the companies that build the most dependable infrastructure will capture the most valuable workloads. If demand is real, the spending is not wasteful. It is a barrier to entry.

The skeptical view is just as simple: there is a long history of technologies that were transformative but temporarily unprofitable for the biggest spenders. In that framing, AI can still reshape the economy while the return on capital shows up slower than the market wants.

Both sides are reacting to the same reality. The spending is happening now. The evidence of payback will arrive unevenly, and investors do not like uneven.

“In the short run, the market is a voting machine but in the long run, it is a weighing machine.”
— Benjamin Graham

How to read the next few months without turning it into stock picks

If you want to follow this story like news, not like a betting slip, focus on signals that are hard to fake.

Start with capacity language. When companies talk about being supply constrained, about doubling data center footprint, or about expanding AI capacity at unprecedented scale, they are usually describing real physical bottlenecks, not marketing.

Then watch whether spending guidance stays stable or keeps creeping upward. A one-time jump can be planned. Serial upward revisions often mean demand is stronger than expected, or that the build is more expensive than expected, and those two scenarios have very different implications.

Finally, track where the first clear monetization shows up. Alphabet has explicitly tied its infrastructure push to revenue and growth, and Meta has tied its step-up to AI priorities while still signaling confidence in operating income. Those are the kinds of statements that become testable over subsequent quarters.

What to watch next

The next phase of this story will not be a single headline. It will be a sequence.

Watch for companies to start breaking out AI infrastructure demand in more concrete ways, whether through backlog commentary, capacity availability, or clearer segmentation inside cloud results. When management teams choose to quantify something, it is usually because they believe the metric will help them, and that alone is information.

Watch the “second-order” constraints. Power, permitting, and grid access are becoming part of the AI narrative, and those frictions can shape where the next buildouts land and how quickly they come online. As the physical world tightens, the returns become more sensitive to execution.

And watch the tone shift from building to sweating assets. Big capex cycles eventually face a simple question: are these data centers getting used at the rate implied by the spending. The market will be listening for early signs of utilization, pricing discipline, and whether AI workloads are sticky enough to defend margins.

If you want a practical way to follow it this year, pick one or two companies and read their capex guidance the way you’d read a weather report. You are not trying to predict the exact temperature. You are watching whether the climate is changing.

The Most Expensive Week to Hesitate

Stacking Trades — Mon, 22 Dec 2025 22:19:54 +0000

The easiest way to misunderstand a missed investment opportunity is to treat it like a single moment you failed to act. You did not buy. You sold too soon. You ignored the headline. That story is comforting because it suggests the fix is simple: next time, click faster.

In reality, the most expensive misses usually don’t occur in a single day. They happen over several weeks when a new advantage is developing, and many people are still unsure if it is genuine. During that time, the advantage has not yet been fully priced, crowded, or normalized. It is something that can be learned. Then the window closes. This doesn’t happen suddenly but through a subtle change in how the market views what is considered the “baseline.”

This article is about that window. It is about the compounding cost of hesitation when a new operating edge appears, especially in AI-driven markets where execution speed is becoming the differentiator.

What actually gets missed

When people say they “missed” an opportunity, they usually mean they missed a price move. But price is the final expression of a deeper change.

The earlier change is operational. A company or a group of companies finds a way to do something faster, cheaper, or more reliably than before. This could involve a new workflow, a new distribution channel, a new feedback loop, or a new way to turn attention into revenue. At first, the improvement seems small and easy to overlook. Then the improvement happens again. This repetition is where compounding starts.

The investment miss is not failing to predict the top performer. The miss is failing to recognize that the operating rules are changing, and therefore the slope of performance is changing.

Why “one week” matters more than it sounds

In slow cycles, a week is just noise. In fast cycles, a week serves as a unit of learning. Teams that adopt early get more attempts. More attempts create more data. More data leads to better decisions. Better decisions result in faster iteration. That iteration appears as improved margins, better conversion, stronger retention, or tighter unit economics. This often happens before it becomes a story that Wall Street shares.

This is what makes hesitation expensive. You are not simply waiting for clarity. You are surrendering attempts. And attempts are the raw material of compounding in modern businesses.

The market tends to price clarity late. It prices “this works” after the people closest to the work have already moved on to “we can scale this.”

The difference between capability and advantage

A common investor mistake in AI-driven markets is to focus on capability, the tool, the model, the feature, the announcement. Capability is easy to see because it ships as a product.

Advantage comes from what a company does with that capability. Advantage is not the tool. Advantage is the workflow that turns the tool into repeatable output. That is why two companies can adopt similar technology and end up with wildly different outcomes. One changes how it operates. The other adds a layer of software on top of the old habits.

This is where many “missed opportunities” actually live. You thought you were waiting to see whether the technology mattered. Meanwhile, the real differentiator was forming at the adoption layer, inside the operating cadence of the business.

The last mile is where the market gap forms

Almost every technology wave has a last mile problem: getting from “it works in a demo” to “it runs the business.”

In AI, that last mile is adoption. It is trust, process redesign, and integration into daily workflows. It is the slow, unglamorous work that makes an advantage durable. It is also where small, well-run teams can outpace larger incumbents because they can change behavior faster.

Sometimes a single case study carries more signal than a dozen product announcements because it shows the last mile in action, what changed in execution, and how the change produced results. One example of that is the way a small budget can start to compound once the process is wired correctly, as shown in RAD Intel’s performance story.

The point is not to anchor to one company. The point is to notice the mechanism: when iteration becomes cheaper and faster, outcomes can scale in a way that looks sudden from the outside.

Why hesitation feels rational at the time

Waiting is not always a mistake. Many investors wait because they are trying to avoid hype and protect capital. That instinct is healthy.

The trap is that the early phase of a real shift rarely looks clean. It looks uneven. It looks like a few outliers. It looks like a narrow group of operators finding leverage while the broader market stays skeptical.

If you require perfect proof, you will usually arrive after the advantage has begun to diffuse. By then, the market narrative is clearer, but the opportunity is often smaller because the learning has already been captured by earlier adopters.

A practical way to read the next window

If you want to avoid missing the next “most expensive week,” stop looking only for headlines and start looking for operational evidence.

The strongest signals tend to be boring. Cycle times drop. Customer response gets faster. A workflow becomes repeatable. A team stops experimenting and starts running a process. Unit economics improve without a matching increase in spend. These are adoption signals, and adoption signals tend to appear before earnings calls and press narratives.

In markets tied to AI, the best indicator is often not that a company has access to powerful tools. It is that the company has built an operating rhythm that turns those tools into throughput.

When you see that rhythm, you are not predicting the future. You are noticing that the present has shifted.

The Algorithm in the Deal Room

Stacking Trades — Mon, 15 Dec 2025 21:52:05 +0000

The modern M&A process starts the same way it always has. Someone believes one company should buy another, and a small group of people works to support that idea. What sets today apart is how quickly these justifications can be put together.

A banker can now pull up a decade of pitch materials, carve out the relevant pieces, and produce something that looks like experience. A diligence team can ask a system to scan thousands of documents for unusual clauses and missing consents, then spend its time arguing about what matters, not where it is. An executive can stress test a synergy thesis in an afternoon, then show up to the next meeting with a confidence that feels earned, even when it is partially borrowed from a model.

AI is creeping into deals the way spreadsheets once did: first as convenience, then as default.

The new front end of dealmaking

The biggest shift is upstream, before anyone signs an NDA. Deal sourcing has always been a mix of relationship and pattern recognition, with a lot of pattern recognition hiding inside relationships. AI is starting to widen that funnel, turning what used to be a whisper network into something closer to a searchable map.

That shift is visible in the tooling vendors and the acquisitions around them. When dealmaking platforms buy “private markets intelligence” businesses that run on AI, they are betting that the next edge comes from getting to the right target earlier, with a clearer picture of where value might be hiding. Datasite’s acquisition of Grata, described as an AI-driven private markets intelligence platform focused on deal sourcing and due diligence, is a clean example of that direction of travel.

The promise is not that an algorithm discovers the perfect target on its own. It is that the long list gets longer, cheaper, and less dependent on who happens to know whom.

Due diligence gets a first draft

The most practical change is in diligence. This work has always been both costly and distinctly human. It involves reading, summarizing, checking, re-reading, and explaining your findings to people who lack the time to read what you read.

McKinsey’s recent work on “outside-in diligence” describes the new workflow plainly: generative AI can take the first pass, synthesizing large volumes of public and proprietary data, identifying trends and outliers, and proposing hypotheses that analysts might not have considered. The caution in that same framing is important too. Many organizations have not yet cracked the operating model that consistently turns tools into impact.

Vendors are racing to make that first pass feel native inside the deal room. Virtual data rooms, historically built for secure sharing and permissions, are adding assistive AI features like automated redaction and review workflows, with the positioning that speed is valuable only if it stays controlled. Datasite, for example, markets AI-enabled tools such as automated redaction, and has pushed newer “Redaction AI” features that still require human review and confirmation.

This is the dealmaker’s version of a change happening across the knowledge economy: AI compresses the time between raw information and a usable narrative. The advantage goes to the team that can validate the narrative fastest, not the team that can generate it.

“In M&A, the winner is often the side that turns information into conviction first.”

Bankers and lawyers become editors of machines

If diligence is where AI changes the muscle, banking and legal teams are where it changes the cadence.

Reuters has reported that Goldman Sachs rolled out a firmwide AI assistant, with around 10,000 employees already using it at the time of the internal memo cited by Reuters. More broadly, large U.S. banks have been describing measurable productivity gains from AI in operations and coding, and executives have been blunt about the implication that doing more with fewer people is now part of the plan.

In M&A work, this means a subtle shift in how labor is distributed. Junior staff still create models and presentations, but their value now lies in verifying, contextualizing, and identifying what the model missed. The focus shifts from creating the initial version to stress testing it. Teams that use AI effectively begin to resemble newsrooms rather than factories. They produce quick drafts, make constant edits, and maintain a strong emphasis on what is true.

Deal structure is changing because regulators are watching

The most underappreciated disruption is not in the spreadsheet. It is in the legal form of the deal.

As AI becomes strategically central, acquisitions and partnerships in the sector have drawn unusually intense scrutiny. The FTC launched a 6(b) inquiry in early 2024 into generative AI investments and partnerships, sending orders to Alphabet, Amazon, Anthropic, Microsoft, and Open AI. In January 2025, the FTC issued a staff report on certain cloud provider and AI developer partnerships, highlighting concerns like lock-in, switching costs, and access to sensitive technical and business information.

That scrutiny has real behavioral consequences. Companies have learned that if a traditional acquisition triggers a fight, there are other ways to get what they want: talent, models, and distribution rights.

Microsoft’s unusual arrangements with Inflection AI became a reference point for this style of “almost acquisition.” Reuters reported that U.S. regulators were looking into the Microsoft Inflection deal in part over concerns it might have been designed to skirt merger disclosure requirements. The UK’s Competition and Markets Authority opened an inquiry and later cleared Microsoft’s hiring of certain former Inflection employees and associated arrangements, treating the situation as a merger inquiry under UK rules even as it ultimately closed the case.

M&A lawyers are already absorbing the lesson: the substance of control, not just the paperwork, is becoming central. The deal is no longer only what you buy. It is how you access models, compute, talent, and data without triggering the harshest form of review.

AI compliance becomes a diligence item

AI is also changing what counts as a material risk.

A decade ago, diligence obsessed over revenue recognition, change-of-control clauses, and litigation exposure. Now, teams increasingly need to understand a target’s model dependencies, data rights, and governance practices. In Europe, this intersects with the EU AI Act’s general-purpose AI obligations, and the voluntary General-Purpose AI Code of Practice published on July 10, 2025, which is positioned as a tool to help providers demonstrate compliance on transparency, copyright, and safety and security.

Even for buyers outside the EU, this matters because global products rarely stay local, and compliance expectations travel through customers, partners, and regulators. In practical terms, AI diligence starts to resemble cyber diligence: less about whether risk exists, more about whether the company has a credible system for managing it.

In the best deals, this does not slow things down. It changes what “fast” means. Fast becomes the ability to answer hard questions with evidence, not the ability to move past them.

AI will not eliminate the most human parts of M&A: the politics, the ego, the leap of faith.

What it will do, and is already doing, is make every stage more legible and more contestable. When both sides have machines that can draft the story, the advantage shifts to the side that can prove it.

The Last Mile of Automation

Stacking Trades — Fri, 12 Dec 2025 19:01:57 +0000

The demo usually looks flawless

A bot copies data from one system to another. A workflow routes a request without the back-and-forth of emails. A dashboard displays clear “time saved” estimates. In the conference room, it seems unavoidable. Then the pilot begins, but people continue to do it the old way. They open the same spreadsheets. They forward the same attachments. The automation is there, but it doesn’t take hold.

If you want to understand why so many automation projects fail, stop staring at the technology. Look at adoption. Look at the tiny, everyday decisions workers make when they are rushing, when they are unsure, when the new system asks for one extra field, when the error message is vague, when there is no clear owner to fix the workflow that broke. That is the last mile. It is also where most programs quietly lose.

The pilot that proves nothing

Enterprise automation has always had a credibility problem: it is easier to automate a process than to automate a company.

Most pilots are created to work well in controlled settings. They choose cooperative users, stable inputs, and a limited scope. The results are not dishonest, but they are weak. When the automation meets the real world, exceptions increase. Edge cases show up. Approvals become political. The data is messier than anyone acknowledged. Suddenly, the system requires humans again, and humans do what they always do under pressure. They find ways to bypass the tool.

This is why “we built it” is not the same as “it works.” The relevant question is whether it changes behavior at scale. McKinsey’s 2025 global survey captures the gap in plain terms: a large share of organizations report using AI in at least one function, but most have not yet scaled the technologies across the enterprise.

The story is similar for automation more broadly. The problem is rarely capability. The problem is absorption.

Adoption is a product problem, not a training problem

When adoption stalls, organizations often reach for the same solutions: more training, more comms, another roadshow. Those help, but they are not the core fix. If a workflow is not being used, assume it is not designed like a product.

Good products minimize cognitive load. They anticipate user intent. They make the next action obvious. They recover gracefully when something goes wrong. In many companies, internal automations are the opposite. They are launched with the mindset of a systems project, not a user experience.

The result is a familiar pattern. The automation creates a new interface, but it does not remove the old one. People now have two ways to do the job, and the old way is still faster when you are experienced, especially when you are dealing with exceptions. Adoption then becomes a social negotiation rather than a natural shift.

This is also where leadership behavior matters more than memos. Recent reporting has emphasized that worker trust and buy-in are now central constraints on rolling out AI and automation, pushing functions like HR and operations into the role of adoption architects rather than policy enforcers.

“Automation doesn’t fail in the lab. It fails in the inbox.”

The myth of the invisible robot

Automation leaders love to say that the best automation is invisible. That is sometimes true for infrastructure, and often false for work.

For most roles, the point is not invisibility. It is reliability and clarity. Workers need to know what the automation did, what it is doing now, and what they are responsible for when something breaks. When that is unclear, automation feels like a black box that can create risk.

This is why “agentic” automation has become such a revealing stress test. It promises autonomy, but it also increases the surface area of uncertainty: what was the agent trying to do, what did it touch, and what happens if it drifts? Gartner has predicted that more than 40% of agentic AI projects will be canceled by the end of 2027, citing issues like rising costs, unclear value, and inadequate risk controls.

Even in the hype cycle, the market is already admitting that the last mile is not just about capability. It is about governance, ownership, and operational fit.

Incentives beat enthusiasm

Adoption fails when the workflow asks people to take on new effort without a clear payoff that is felt immediately.

A sales team will not use a new automation if it adds steps before a deal can move forward. A support team will not trust an automated routing system if it occasionally sends high priority tickets into a void. A finance team will not rely on a bot that cannot explain why an invoice was flagged. In each case, the rational choice is to build a parallel manual process “just in case,” and that parallel process quietly becomes the real one.

The deeper issue is incentives. Many automation programs measure success by output metrics, how many workflows were built, how many hours were “saved” on paper, how many bots are in production. Those numbers can look great while adoption is flat. The incentives reward shipping, not usage.

When the metric becomes adoption, the program changes shape. Rollouts become slower and more iterative. Exceptions become the main product. Documentation stops being an afterthought. Owners get named, not as governance theater, but as the people who will respond when the workflow fails at 4:55 p.m.

From project to product, the only move that scales

One of the most consistent observations in recent management discussions is that initiatives fail when organizations are not set up to support them. Harvard Business Review has stated this clearly in relation to AI. Failures often arise not from weak models, but from companies lacking the structure, operating rhythm, and accountability needed to maintain systems effectively after launch.

The best automation programs look less like implementations and more like product lines. They have backlogs driven by real user pain. They ship small improvements continuously. They treat governance as part of design rather than as a gate at the end. They invest in measurement frameworks that track workflow outcomes, not just activity.

This mindset also counters a newer failure mode: transformation fatigue, the exhaustion that sets in after too many top-down tools arrive with big promises and small practical value. When workers have lived through enough underwhelming change, adoption stops being a tool-by-tool decision and becomes a cultural reflex: wait it out.

If the last decade of automation taught enterprises how to build, the next one will teach them how to land.

The Developer Experience Economy

Stacking Trades — Wed, 10 Dec 2025 21:32:35 +0000

The first sign that something had changed was not a new programming language or a popular open source library. It was a slide in a board meeting. Alongside revenue, margins, and churn, a fourth chart showed up: deployment frequency and DevEx score. The message was clear. How developers felt about their tools had become a business metric.

For years, companies treated internal developer experience as a kind of housekeeping, important but rarely urgent. In 2025, it has become a competitive weapon. Research from Google and independent DevEx labs now treats productivity as a function of speed, ease, and quality, measured through a mix of telemetry and direct surveys. Gartner tracks a growing market for “internal developer portals,” and consulting firms sell playbooks for unlocking revenue growth through happier engineers. What used to be tickets in a backlog is now a line in the strategy memo.

When productivity became a product

The shift starts with a simple arithmetic problem. In the 2024 Stack Overflow Developer Survey, a majority of professional developers reported spending more than thirty minutes every day just searching for answers to work problems. That is time spent in documentation mazes, chat histories, and half-remembered Confluence pages rather than in the codebase.

At the same time, the average toolchain has grown more complex. A developer working on a single feature might interact with the source repository, a feature flag service, a build pipeline, a cloud console, an observability platform, and several chat channels before the change reaches production. Each step adds friction. Each missing script or unclear error message adds a small cost.

The result is that developer experience itself has started to look like a product surface. Companies now build internal platforms with the same care they once reserved for customer facing apps: user research, design reviews, roadmaps, and service-level objectives tailored to engineers.

The rise of the internal developer platform

Platform engineering emerged as a response to this sprawl. Rather than ask every team to stitch together its own path to production, organizations began building centralized “golden paths” that abstract away infrastructure and policy decisions. The idea is not new, but the tooling is.

Open source frameworks like Backstage, created at Spotify and donated to the Cloud Native Computing Foundation, turned the concept of an internal portal into reusable software. Backstage catalogs services, pipelines, and documentation in one place, so engineers can discover what exists and scaffold new projects with consistent templates. A growing ecosystem of SaaS platforms now wraps these ideas in managed offerings, promising faster onboarding and reliable standards without the pain of building everything in house.

Analysts have started to quantify the trend. Gartner defines internal developer portals as the front door to reusable components, tools, and knowledge, and projects that by 2028 most organizations with platform engineering teams will offer one, up from about sixty percent in 2025. What was once a niche initiative now looks like the default infrastructure pattern for serious software companies.

“Developer experience used to be what was left over after the tools were chosen. Now it is the thing being designed.”

Metrics move into the boardroom

Once internal platforms are in place, leaders want to know if they are effective. This has led to a surge of frameworks that aim to measure not just lines of code or tickets closed; they also focus on the real experience of delivering software.

Google’s productivity researchers have described an approach that blends survey data with system metrics to capture speed and ease in a way that developers recognize as real. Academic and industry teams have published frameworks such as SPACE and DevEx that frame productivity across satisfaction, performance, communication, and flow. More recently, the Core 4 model has tried to unify these ideas into a concise set of outcomes that leadership can track without turning engineers into KPI collectors.

Consultancies and vendors have leaned into the story. Deloitte describes DevEx as a lever for product innovation and operational efficiency. Studies aggregated by DevEx tooling companies suggest that organizations investing in better developer workflows see significantly faster time to market and improved customer acquisition. HashiCorp cites McKinsey research that links strong developer experience to higher operating margins. The numbers are imperfect, but they share a direction. Developer friction is now described not as “annoying” but as a drag on revenue.

AI tools and the new bottleneck

The timing is not accidental. As AI coding assistants spread through the industry, the cost of writing syntax has fallen. According to recent survey data, more than eighty percent of developers now use or plan to use AI tools in their workflows, yet many still spend large chunks of the day hunting for context and debugging almost-right suggestions. The bottleneck has shifted from typing code to orchestrating the environment in which that code runs.

That is where internal platforms matter. An AI tool can write a microservice, but the organization still needs an opinionated way to connect that service to authentication, observability, and deployment. A cluttered CI system or inconsistent staging environment can erase the gains of even the best assistive model.

The companies that benefit most from AI coding tools are often the ones that already invested in clean paths to production. When a scaffolded project comes with batteries included, an AI agent can safely generate more of it. Developer experience becomes the substrate that makes automation trustworthy instead of chaotic.

From perk to strategy

For a long time, improving developer experience was framed as a retention play. The logic was that happier engineers were less likely to leave, and perks like better laptops or cleaner tooling were part of that equation. That lens has not gone away, but it is no longer sufficient.

What is changing is that DevEx is being folded into the core story of how a company competes. Internal portals, standardized workflows, and thoughtful documentation become part of the answer to investor questions about how a business will ship new products faster than rivals. Platform teams are judged not only on internal satisfaction scores but on their contribution to time to market and stability.

The organizations that treat developer experience as an economy of its own are starting to look different inside. Projects spin up with fewer meetings. New hires find their footing in days instead of weeks. AI tools amplify good patterns instead of copying bad ones. The work of building internal tools and platforms is still largely invisible to customers, but its effects are not.

In an era where the external technology frontier is moving quickly, the real differentiator is often what happens inside the walls of a company. The developer experience economy is the quiet infrastructure behind that edge, turning the messy, improvised workflows of the last decade into something more deliberate, measurable, and, increasingly, strategic.