PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWhen software takes less effort to build, more projects may become worth attempting and teams may be able to do more with the same labor. But cheaper code does not automatically mean cheaper reliable software: deciding what to build, checking it, integrating it, securing it and maintaining it still takes work. The evidence so far measures different things—from software prices to task completion—and does not establish a universal effect on software demand or developer jobs.
What does “cheaper to build” mean?
Software has several costs, and they do not necessarily move together. A team may spend less time producing code while spending as much or more on requirements, review, testing, deployment, support and future changes. A faster first version is not the same as a lower cost over the software’s full life.
It helps to distinguish at least four measures: the price of software sold in the market; labor or time spent writing code; work completed by a team; and useful software delivered and kept running. Evidence about one measure cannot automatically answer questions about the others.
What evidence shows about software prices and coding productivity
The available findings point in different directions because they examine different populations, tasks and outcomes. They should not be treated as a direct ranking of tools or a forecast of savings for a particular company.
Recommended Free Tools
#1 Best Overall
| Evidence | What was measured | Reported result and scope |
|---|---|---|
| BEA-hosted paper (2024) | Software price changes from 2015–2021, using an alternative measurement method and comparing it with the published NIPA measure. | The paper estimates annual price declines of 6.4% under its method, versus 2.0% in the published measure. This is a measurement of software prices, not a universal estimate of the labor or lifecycle cost of building bespoke software. |
| Three company field experiments summarized by Microsoft Research (2025) | Completed tasks among 4,867 developers at Microsoft, Accenture and an anonymous Fortune 100 company, comparing developers offered an AI coding assistant with the relevant control groups. | The combined result was a 26.08% increase in completed tasks, with a 10.3% standard error. It is a result from these experiments, not a guaranteed effect across organizations or tasks. |
| METR randomized study (2025) | 246 tasks undertaken by 16 experienced developers working in their own mature open-source repositories with early-2025 AI tools. | Developers took 19% longer on average with the tools in this study. Its small, specific sample makes it a caution against assuming that results transfer to other developers, tools or work settings. |
| GitHub controlled experiment, reported in 2023 and updated in 2024 | A 2022 task in which developers implemented a JavaScript HTTP server, with Copilot compared with a control group. | The Copilot group completed this task 55.8% faster. That narrow task result does not establish a comparable reduction in the time or cost of a whole software project. |
| NBER Working Paper 35275 (2026) | An analysis using data from more than 500,000 GitHub developers, reporting estimates at several output levels. | The reported estimated effect attenuates from 240% for code to 80% for projects and 30% for releases. It is a working-paper estimate, and the gap between these measures illustrates why code production is not interchangeable with shipped outcomes. |
These results are not inherently contradictory. A short, well-defined task or a company experiment can produce a different result from work inside a mature repository, where an experienced developer must understand existing conventions and avoid regressions. The relevant question is not simply whether a tool helps, but which work it helps with and what additional effort is needed to turn its output into a dependable release.
Where the work can move when code takes less effort
Lower implementation effort can change the balance of work without eliminating the rest of software development. In many projects, the consequential work shifts toward choosing valuable problems, defining expected behavior, checking proposed changes and supporting the system after launch. This is a useful way to reason about the change, not a measured rule that applies to every team.
- Problem selection: Teams can consider smaller or more specialized ideas that previously did not justify the initial build effort. Whether those ideas deserve investment still depends on user need, business value and competing priorities.
- Specification: Developers and product teams still need to decide what a feature should do, how it handles edge cases and what it must not do. Ambiguous requirements can produce more code without producing a more useful result.
- Review and integration: Generated or quickly written changes must fit the existing system, work with other changes and remain understandable to the people responsible for them.
- Validation: Tests, security checks, performance checks and reliability work remain important wherever a failure could harm users or disrupt operations. Less time spent drafting code does not itself show that these risks have fallen.
- Operations and maintenance: Software must be monitored, repaired and adapted as dependencies, infrastructure and user needs change. A lower initial implementation cost does not establish a lower cost to operate or maintain it.
The practical consequence is that a team may start more experiments or prototypes, but the bottleneck can move to deciding which ones to finish and ensuring that finished work meets its requirements.
Why more code does not necessarily mean more software value
Output can be counted at different levels: code written, tasks completed, projects started and releases shipped. Each step adds requirements that the prior measure misses. Code can be produced without completing a task; a completed task may not make a coherent project; and a project may never become a release that users can rely on.
The NBER working paper’s reported estimates, which decline across code, projects and releases, make that distinction especially important. For an organization, the useful measure is the outcome it actually needs—such as a tested feature released to users—not a proxy that is easier to count. Review time, defects, security findings, support burden and maintenance effort can help explain whether faster production translated into a net improvement.
Does widespread AI use prove that software is cheaper?
No. GitHub’s 2024 survey of 2,000 enterprise software-team respondents in the United States, Brazil, Germany and India found that more than 97% had used generative AI tools at some point. The survey was fielded in February and March 2024 and measures respondents’ self-reported exposure. It does not show that 97% of firms formally approved the tools, embedded them in everyday workflows or realized savings.
Rank #4
Adoption can show that a tool has reached developers; it does not establish how much total work it displaces or whether the resulting software is better, cheaper to operate or more valuable. Those outcomes require measurement at the team and product level.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What cheaper software could change—and what is not established
As economic reasoning, lower build effort could make more software ideas viable, enable organizations to tailor tools to narrower needs, or let existing teams deliver more with a fixed amount of labor. If supply expands, prices could also change. These are possible channels, not conclusions established by the findings above.
Best Value
The evidence here does not settle whether lower building costs will increase total software demand, reduce market prices across the board, encourage more firms to form or reduce software employment. Hiring depends on more than the cost of writing code: demand for software, the volume of work organizations choose to pursue, required oversight and maintenance, and the skills needed to deliver dependable systems can all matter. A productivity gain in one kind of task is not, by itself, evidence of a particular employment outcome.
How to tell whether a team is actually benefiting
A useful evaluation compares like with like and follows work through release rather than stopping at code generation. Before adopting a tool or changing a workflow, define the task category and the outcome that matters; then track the costs and quality checks needed to reach that outcome.
- Choose representative work. Separate well-specified, routine tasks from unfamiliar or high-risk changes in mature systems. Results from one category should not be assumed to apply to another.
- Record the full effort. Include time spent prompting or setting up a tool, reviewing changes, fixing errors, integrating code and maintaining the resulting feature—not only time to produce an initial draft.
- Use an outcome that matches the goal. If the aim is delivery, measure accepted work or releases. If the aim is quality or reliability, include the relevant tests, defects or operational measures rather than treating volume of code as success.
- Compare under consistent conditions. Account for developer experience, familiarity with the codebase, task complexity and tool context. A small controlled test can inform a local decision but should not be presented as a universal productivity rate.
- Keep the safeguards in view. Check security, reliability and maintainability as part of the result. A faster implementation that raises downstream risk may not reduce the cost that matters.
Software becoming cheaper to build would therefore mean more than writing code faster. The durable change would be a lower total effort to deliver and sustain useful software. The available evidence shows that some measured prices or task outcomes can move substantially, but it does not show that every project—or the full lifecycle—gets cheaper by the same amount.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
