The PDF is a postmortem on Direct File, the IRS service that let some taxpayers file federal returns directly with the government for free instead of going through TurboTax, H&R Block, or another intermediary. The report argues that Direct File shipped unusually fast for a federal product, got strong user satisfaction, and proved that a government team could deliver competent software under ugly constraints. What grabbed people was not whether the web app itself worked. Most took that as settled. The live question was why a seemingly successful public service was still so easy to kill.
The strongest reading was structural, not technical. Direct File sat in a bad political lane from day one. Congress asked for a study, not a clearly protected permanent program. Federal procurement culture also leans hard toward buying from contractors instead of building inside the agency. That made Direct File vulnerable even before partisan turnover. Several people connected its fate to the tax prep industry’s long campaign to preserve filing complexity and keep the IRS from becoming a direct channel. Others broadened that into a more general point about US governance right now. A working program is not safe if a new administration can reverse it faster than Congress can defend it.
The cost debate mostly landed in the same place. Per-return numbers from the pilot were treated as a bad way to judge it because they bundle one-time build cost with a deliberately small rollout. The more useful comparison is fixed build cost versus ongoing operating cost, plus the time and fees taxpayers avoid when they stop re-entering data the IRS already has. A smaller but persistent objection was legal and institutional. If Congress only authorized a study, then a live filing product overstepped. A few people also distrusted the basic model of the tax collector helping prepare the return, arguing that taxpayers need an advocate whose incentives run the other way. That view did not carry the room. Most saw private tax prep as the more obvious conflict of interest, since those firms profit from complexity and from standing between taxpayers and the government.
If you build products that depend on government policy, treat operational success as only one layer of risk. The harder problem is durability through procurement rules, budget ownership, and changes in political leadership.
Mostly frustrated and cynical. People broadly viewed Direct File as a competent, user-friendly government product that was killed for political and lobbying reasons rather than product failure, with irritation also directed at procurement rules that favor contractors and at bad-faith cost comparisons from a limited pilot.
Key insights
01
Conflict of interest cuts both ways
The sharpest substantive pushback was not about code quality or even cost. It was about incentives. One view held that the IRS should not both collect taxes and help compute what you owe, because taxpayers sometimes need an advocate to claim deductions or credits aggressively within the law. The answer that landed harder was that this risk is limited for simple returns, which were the actual target, and that private preparers have their own conflict because they make money from complexity and from charging low-income filers to recover refunds the government already has most of the data to estimate.
Separate simple returns from edge cases in your own thinking. A direct public workflow looks much stronger when it is scoped to routine cases and leaves room for paid advisors where judgment calls actually matter.
Procurement rules stacked the deck against in-house software
A government vendor argued that Direct File was always politically fragile because federal policy has favored "buy" over "build" for decades. That matters more than the usual red-blue story. Agencies are pushed toward commercial contractors, system integrators, and inter-agency delivery models that often leave the customer agency holding maintenance risk later. The useful read here is that Direct File did not just threaten TurboTax. It also cut across a procurement ecosystem that expects software money to flow through vendors, contract vehicles, and billable-hour structures.
If you are selling into government or building internally for a regulated buyer, map the procurement incentives as carefully as the user journey. A product can win on usability and still lose because it bypasses the channels institutions are built to fund.
The most useful correction to the "$226 per return" line was that it mixes startup spending with a small pilot denominator. That makes the number look like a recurring per-user cost when it is mostly a first-build cost. The better frame is infrastructure economics. You pay heavily to create the system, then marginal cost drops as usage grows. Several people added that a fair comparison should also count taxpayer time, data re-entry, and filing fees. A pure government-versus-market sticker-price comparison misses the cost of forcing millions of people through intermediaries.
When you evaluate a new internal platform or public digital service, separate fixed build cost from marginal operating cost before making a unit-economics call. Also price the user’s time if the alternative is a fragmented vendor workflow.
The report’s most durable lesson was not that government can ship software. Plenty of countries already know that. It was that a functioning service can remain institutionally weak if it lives as an executive initiative instead of as a clearly protected statutory program. Even a bipartisan law is not a perfect shield, as commenters pointed to other recent cases where administrations simply stopped enforcing or actively unwound mandated programs. Success at launch did not solve the continuity problem.
For any policy-adjacent product, ask early what survives a leadership change and what exists only by permission. If continuity matters, plan for legal and organizational entrenchment, not just adoption metrics.
The existing filing stack is deliberately confusing
One useful detail was that many people do not understand what is actually operated by the IRS versus by the Free File Alliance and commercial vendors. That confusion is not accidental. Commenters described a long-running compromise where free filing existed in a way that preserved private intermediaries and kept the direct government path obscure or burdened with extra hoops. Direct File mattered partly because it cut through that maze and made the filing relationship legible again.
When incumbents defend a market through complexity, a cleaner default path can be more disruptive than a technically superior product. Clarity itself becomes a competitive threat.
A credible minority argument was that Direct File’s policy weakness was not just politics but process. Congress funded a study, not a standing service, and the IRS moved from evaluation into a live product used by taxpayers. The rebuttal was that building and testing a real prototype is exactly how you study adoption and feasibility. Still, if you care about administrative legitimacy, this is a real vulnerability rather than a talking point.
Do not assume a pilot label protects an ambitious rollout. If your program depends on a narrow authorization, document exactly how live use fits the mandate before success creates a bigger target.
Distrust of the IRS is grounded in lived experience
The anti-Direct File case was easy to dismiss when stated abstractly, but it drew force from personal stories about audits, arbitrary enforcement, and opaque contractor behavior around tax compliance checks. That does not prove Direct File would worsen any of it. It does explain why some people reject a government-prepared return on principle. For them, the issue is not convenience. It is whether the same institution that can penalize you should also be the default interpreter of what you owe.
If a public digital service depends on trust in an unpopular agency, usability alone will not carry it. Adoption and durability improve when people can inspect the calculation, opt out easily, and verify decisions independently.
ischool.berkeley.edu