Structured from 9 source documents across 3 independent research lenses (generational turnings, historical precedent, AI impact). Source-reviewed, fact-reviewed, and gap-reviewed before publication.
There is a theory — contested, unorthodox, rejected by most professional historians — that says institutional trust follows an eighty-year rhythm. It rises after a crisis, peaks during a postwar High, erodes through an Awakening, collapses through an Unraveling, and then the Crisis forces new institutions into existence. William Strauss and Neil Howe first proposed this cyclical model in The Fourth Turning (1997), and Howe updated it with contemporary evidence in The Fourth Turning Is Here (2023). They call this final, crisis phase a Fourth Turning. The last one peaked in the early 1930s, when trust in banks, government, and business had all collapsed simultaneously — thousands of bank failures, Hoover’s paralysis, the Pecora Commission exposing Wall Street fraud. The resolution wasn’t a return to normal. It was the New Deal, the FDIC, Social Security, the Wagner Act — entirely new institutional arrangements built under crisis pressure.
Right now, by every longitudinal measure we have, institutional trust is at or near that level again. And the evidence is not subtle.
But before looking at the data, consider what happened the last time a major profession was reorganized by technology and the government tried to help.
The most effective government retraining program for displaced workers in the United States — Trade Adjustment Assistance, the primary federal response to offshoring — produced outcomes worse than doing nothing at all. Not marginally worse. Measurably, consistently worse. Brookings found that TAA participants had lower employment in the first years after layoffs compared to similar workers not in the program. Reynolds & Palatucci (2012) quantified it: participating in TAA caused a wage loss approximately ten percentage points greater than not participating.
That is not a funding problem. It is not an implementation gap. It is a structural finding about what institutions can and cannot do when a profession is being reorganized by technology. And if the cyclical pattern is even approximately right about where we are, it is exactly what you would expect: institutions failing at the precise moment they are most needed, because the old institutional arrangements are dying and the new ones have not yet been forced into existence.
The arc of the broken promise
The relationship between workers and institutions — employers, government, unions — has been eroding for decades. But it is useful to see the arc whole, because the current moment is not a dip. It is a terminus.
In 1970, 45% of private-sector workers in the United States had defined-benefit pension plans — the kind where the employer bears the risk and the worker gets a guaranteed income in retirement. The Revenue Act of 1978 introduced the 401(k), and companies began shifting retirement risk from themselves to their workers. By 2025, only 13% of private-sector workers have traditional pensions. That is not a policy adjustment. It is a forty-year transfer of financial risk from the institution that can bear it to the individual who cannot.
The same transfer happened with career development. Companies once invested in training, mentorship, internal career ladders — not out of benevolence, but because lifetime employment norms made it profitable to develop the workforce you intended to keep. That model is gone. Employers now expect workers to arrive “job-ready” and treat them as interchangeable. The rational response is already visible in career behavior: employees who switch companies earn an average salary increase of 7.7% versus 5.3% for those who stay. Only 17% of workers are promoted internally over five years, while 58% change jobs to advance.
Then came the phase that made the implicit message explicit.
In 2025, at least ~127,000 tech workers were laid off across hundreds of companies, with another 52,150 across 146 companies in early 2026, according to TrueUp’s layoff tracker. These were not struggling companies cutting costs to survive. Microsoft reported strong sales and profits beating expectations in the same quarter it conducted mass layoffs. The average S&P 500 CEO earned $18.9 million in 2024, putting the CEO-to-worker pay ratio at approximately 285-to-1. When a company lays off workers during a record profit quarter while increasing executive compensation, the implicit message becomes unmistakable: labor is a cost to optimize, not a relationship to maintain.
Workers have received this message clearly. Glassdoor data shows “misalignment” mentions in employee reviews up 149% and “distrust” up 26%. Layoff mentions in late 2025 reviews exceeded March 2020 pandemic levels. Small layoffs — under fifty people — rose from 38% of all layoffs in 2015 to 51% in 2025. Companies learned that rolling cuts under the headline threshold avoid public backlash while continuously eroding workforce stability.
The arc from pensions to gig economy to rolling layoffs during record profits is not a series of unrelated events. It is a single trajectory: the systematic transfer of risk, cost, and uncertainty from institutions to individuals. Each step was individually rational for the institution. The cumulative effect is that the implicit social contract — loyalty in exchange for stability, development, and honest communication about the future — is broken.
The trust collapse by the numbers
The subjective feeling of broken trust is now confirmed by every major survey instrument with decades of longitudinal data.
Pew Research tracks government trust across nearly seven decades. The share of Americans saying they trust the federal government “always” or “most of the time” has not exceeded 30% since 2007. It currently sits near a 70-year low.
Gallup’s Confidence in Institutions survey, tracking nine core institutions since 1979, shows average confidence at 28%. Only three institutions command majority confidence: small business (70%), the military (62%), and science (61%). Congress and television news sit at roughly 10%. Large technology companies scored 26%. Big business scored 16% — near its lowest recorded level. The partisan gap in institutional confidence is the largest in 46 years of measurement.
The 2026 Edelman Trust Barometer, surveying nearly 34,000 respondents across 28 countries, found that grievance has devolved into insularity — 70% of respondents are unwilling or hesitant to trust someone with different values, backgrounds, or information sources. Only 32% globally believe the next generation will be better off. In the United States, that figure is 21%, down nine points. The income-based trust gap has doubled from 6 points in 2012 to 15 points in 2026, with the largest disparity in the U.S. at 29 points.
Trust in the technology sector specifically has collapsed. Eight years ago, technology was the most trusted sector in 90% of countries tracked by Edelman. Now it is most trusted in only half. In the U.S., trust in AI companies fell from 50% to 35% over five years.
The Urban Institute documented a 22-percentage-point decline in trust across major U.S. institutions since 1979, with the “informed public” 18 points more trusting than the general population. Between 2012 and 2021, the number of countries with double-digit trust inequality tripled.
These are not opinion polls about whether people are having a good year. These are longitudinal instruments tracking something structural: the population’s assessment of whether institutions function as they claim to. The assessment is the lowest it has been in the lifetimes of most working professionals.
The pattern underneath the numbers
Map the trust data onto the Strauss-Howe cycle and the shape comes into focus. During the postwar High (1946-1964), 77% of Americans trusted the federal government — institutions were strong, civic confidence was rising, the GI Bill and corporate career ladders actually delivered. Through the Awakening (1960s-1980s), trust began eroding as people questioned institutional values. Through the Unraveling (1984-2008), cynicism became the default posture and institutions lost legitimacy. And now, in the Crisis — the Fourth Turning: Pew at a 70-year low, Gallup at 28%, Edelman documenting a society that has turned from grievance to insularity.
This is not a dip that rebounds. In the cyclical framework, this is structural: the phase where old institutions lose their mediating function and have not yet been replaced. The 1932-1933 parallel is not metaphorical. The shape of the data is the same — maximum institutional discreditation, arriving right on schedule eighty years after the last time it happened.
If the framework is even approximately right about timing, developers are making career decisions in the equivalent of 1937-1941 — past the worst of the initial crisis, deep into what Howe calls the consolidation phase, but several years away from the climax that will force total reallocation of resources and labor. Neil Howe projects that climax — the ekpyrosis, where “whole institutions are killed and new institutions are born overnight” — around 2028-2032. The observable data as of early 2026 is broadly consistent with that timeline. Peter Turchin’s structural-demographic theory, a separate peer-reviewed framework with stronger empirical credentials, independently predicted in 2010 that the US would enter peak instability in the 2020s. Two methodologically distinct frameworks converging on similar timing does not prove either one right. But it does suggest the macro pattern is real, even if the mechanism is debated.
You do not need to accept cyclical theory to find the trust data alarming. But the framework offers something the numbers alone do not: a reason to believe this is not random, that the rhythm has a resolution, and that the resolution looks nothing like a return to the way things were.
The developer-specific trust gap
Within the broader collapse, developers have their own version — and it is particularly striking because developers are the people building the tools that their own employers claim will transform everything.
The 2025 Stack Overflow Developer Survey — approximately 49,000 respondents — mapped the gap precisely. Trust in AI tool accuracy fell 11 points in a single year, from 40% to 29%. Nearly half of all developers — 46%, up from 31% — now actively distrust the accuracy of the tools their employers are deploying. This is not a sentiment dip. It is the developer profession’s version of the Gallup confidence collapse, playing out on a compressed timeline.
The organizational dimension is sharper. Deloitte’s TrustID Index found that trust in company-provided generative AI fell 31% between May and July 2025 — a two-month collapse, not a gradual erosion. Trust in agentic AI systems dropped 89% in the same period. The people with the deepest technical judgment — senior developers who architect the systems these tools are supposed to improve — are the biggest holdouts on AI adoption. That is not resistance to change. It is institutional trust failing at the point where expertise meets corporate strategy.
The developers are not wrong. METR’s randomized controlled trial found experienced developers completed tasks slower, not faster, with AI assistance. CodeRabbit’s analysis of pull request quality found AI-generated code introducing significantly more defects per review cycle. The profession has run the experiments the marketing materials cite. The experiments contradict the marketing materials. And yet the deployment continues — because the deployment decision sits with executives, not engineers.
When a workforce overwhelmingly adopts a tool under institutional pressure while overwhelmingly distrusting its output, the word for that is not adoption. It is compliance.
What the historical record shows about institutional help
If trust in employers is broken and trust in government is at a generational low, perhaps organized labor fills the gap. Or perhaps government retraining catches up. Or perhaps the next administration gets it right.
The historical record on all three is unambiguous.
Government retraining: TAA is the example that has the most data, and the data is not kind. Beyond the Brookings findings on short-term negative outcomes, the program initially excluded IT workers entirely because software was not classified as a “tangible good.” It took until 2009 — nearly a decade into the offshoring wave — for Congress to extend eligibility to tech workers. An estimated 155,000 workers who would otherwise have been ineligible gained access after the change. By then, the sharpest displacement was over. The Job Training Partnership Act showed no statistically significant improvement in employment outcomes. The Workforce Innovation and Opportunity Act has not produced results meaningfully different from its predecessors.
The pattern is consistent: government retraining arrives late, targets broadly rather than specifically, and produces outcomes that are negligible to negative in the short term. The Hyman (2018) finding that TAA participants had roughly $50,000 in higher cumulative earnings over a decade sounds better — until you learn the advantage declined to near zero by year ten. The training did not set participants on a higher trajectory. It temporarily buffered the fall, then converged with the counterfactual.
Meanwhile, the most effective transition mechanism Brookings identified was on-the-job learning in a new role — not formal retraining, but getting hired into something different and learning by doing. This consistently outperformed every program because skills were immediately applied to actual job performance. But that mechanism requires getting the new role first. That is not a program. That is the challenge itself.
Union resistance: Unions have negotiated financial cushions — the longshoremen’s container royalty, the New York Times’s lifetime employment guarantees for 800+ typesetters — but they have not prevented displacement in any case studied. More critically, resistance to the technology itself has consistently backfired. The International Typographical Union’s 1962-63 New York newspaper strike — 114 days — accelerated the industry’s decline. Four of the seven affected newspapers eventually ceased independent publication. At Wapping in 1986, Rupert Murdoch fired 6,000 print workers overnight when they struck against computerization. Approximately 5,000 lost their jobs permanently.
The ILA’s 2024-25 automation ban — blocking automated container terminals — bought temporary delay, but the agreement is explicitly temporary, expiring in 2030. History suggests the delay will not prevent the displacement; it will only determine whether the transition is managed or abrupt. And software development has no equivalent union. No collective bargaining. No hiring hall. No mechanism for the profession to negotiate its own transition terms.
Corporate retraining: Companies that offshored rarely invested in retraining displaced domestic workers. The savings went to shareholders and consumers. The current AI cycle shows the same pattern: companies are investing aggressively in AI tooling while conducting layoffs. The investments are in the technology that displaces the workers, not in the workers the technology displaces.
The single success — and why it doesn’t transfer
There is one documented success in the entire dataset. Sweden’s job security councils — employer-funded, jointly managed transition organizations — achieve approximately 90% reemployment rates for displaced workers, with 34% matching or exceeding prior wages.
The mechanism works because the institutional infrastructure exists before the displacement occurs. Employers fund the councils. Unions and employer associations jointly govern them. Counselors provide individualized career support — not classroom retraining, but direct job matching and on-the-job transition assistance. The system works because it is embedded in a labor market where all parties have agreed, in advance, to share the cost of managing transitions.
The United States has no equivalent infrastructure. No employer-funded transition councils. No jointly governed career support. No agreement between labor and capital to share the costs of displacement. Building this infrastructure would require political consensus, employer buy-in, and years of institutional development — none of which is visible on any horizon. The Swedish model is proof that institutional help can work. It is simultaneously proof that it requires institutional capacity the US tech sector does not have and is not building.
The counterargument: your employer might be fine
There is an important counterpoint that the macro data alone obscures, and intellectual honesty requires stating it directly.
The 2026 Edelman Trust Barometer found that “my employer” sits at 78% trust — the highest of any institution measured. Not government at 17%. Not big business at 16%. Not the technology sector at 35%. Your specific employer, at 78%. Seventy-four percent of high-trust employees intend to remain loyal, and 83% say they are strongly committed to their jobs.
This is not a portrait of a universally broken relationship. It is a portrait of a bifurcated workforce. Workers who have experienced layoffs, AI-driven displacement, or visible CEO-to-worker pay disparity show collapsed trust. Workers who have not yet been directly affected maintain relatively high employer trust — and for them, that trust may be rational and well-calibrated.
The “treat your employer as a strategic partner, not a family” posture is not a universal prescription. A developer at a firm with genuine profit-sharing, transparent AI communication, and stable headcount faces a different decision calculus than one at a company conducting rolling silent layoffs. The macro collapse is real, but its distribution is uneven.
The diagnostic is observable. Your employer has conducted layoffs in the last eighteen months while reporting profit growth. Executive compensation has increased while headcount has decreased. AI is cited in earnings calls but not discussed transparently with engineering teams. Internal career development programs have been cut or frozen. Your role’s scope has expanded without a corresponding title or compensation adjustment.
The more of those that are true, the more the macro-level trust collapse is a live variable in your individual career. The fewer that are true, the more your situation may resemble the high-trust end of the bifurcated workforce. Assess the specific institution, not the category.
The structural absence
The deeper problem is not that institutions are failing to help. It is that the kind of institution capable of managing this transition does not exist for software developers.
The switchboard operator transition was managed by AT&T — a monopoly employer that deliberately chose to go slow. The company viewed human operators as central to service quality and managed the pace of automation through natural attrition rather than mass layoffs. But AT&T could do this because it was a regulated monopoly with no competitive pressure to automate faster. No software company occupies that position. Microsoft, Google, and Amazon control the AI tools, and they are accelerating adoption as fast as possible. They are not managing the pace of disruption; they are setting it.
The agricultural transition benefited from decades of institutional development — the Extension Service, land-grant universities, the GI Bill, industrial employment absorbing displaced farm labor. The typesetter transition had the ITU, which ultimately failed but at least negotiated financial buffers for some workers. The longshoremen had the ILA and ILWU, which extracted meaningful concessions — container royalties, guaranteed employment for existing workers — even though they could not prevent 90% job losses.
Software development has no monopoly employer managing pace. No dominant union negotiating terms. No government body coordinating the transition. No Extension Service teaching displaced developers new skills in adjacent fields. Every company is making independent adoption decisions simultaneously, each one optimizing for its own competitive position. The aggregate result is that no one is optimizing for the profession.
The Strauss-Howe framework explains why this void exists at this particular moment. During the consolidation phase of a Fourth Turning, old institutional arrangements are actively dying but the crisis has not yet forced the construction of replacements. The New Deal agencies that managed the Depression-era transition did not exist in 1931. The GI Bill that retrained millions after WWII was not signed until 1944. The institutions that will manage the AI transition — whatever they turn out to be — do not exist yet because the crisis has not yet reached the intensity that forces their creation. We are in the gap between institutional death and institutional birth. That gap is the structural reason there is no cavalry.
Meanwhile, government capacity is actively shrinking. DOGE eliminated approximately 212,000 federal workers and terminated an estimated $5-7 billion in annual IT contract value. The entity most likely to coordinate a professional transition is reducing its own workforce and its own demand for software development. This is not an abstract concern about future government capacity. It is a current reduction in the size of the developer labor market.
The only reliable strategy
The evidence points in one direction: self-directed adaptation is the only reliable strategy. Not because individual grit is superior to institutional support — Sweden’s job security councils prove otherwise — but because the institutional support does not exist, is not being built, and the institutions that might build it are themselves in crisis.
What self-direction looks like, concretely:
Build career capital that survives institutional failure. Open-source contributions, public technical writing, cross-domain expertise, professional networks outside your employer, financial reserves that give you the ability to walk away. The pension-to-gig-economy arc is a forty-year lesson in what happens when you let a single institution hold your professional future. Don’t let it hold yours.
Build your own AI understanding. Trust in company-provided AI fell 31% in two months for a reason. Your employer’s AI strategy is shaped by executive narrative needs, not engineering reality. The senior developers who are most skeptical of AI tools are also the ones with the deepest technical judgment. Build your understanding from the primary evidence — the METR study, the Faros AI findings, the actual capabilities and limitations you can test yourself — not from your employer’s internal marketing.
Assess your specific employer, not the macro trend. The 78% figure for employer trust is real. If your company is genuinely investing in its people, communicating transparently about AI’s role, and not conducting layoffs during profit growth — that institutional relationship has value. Don’t abandon it for a generalized narrative of institutional collapse. But don’t confuse a good current employer with a permanent guarantee either. The signals are observable. Watch the behavior, not the mission statement.
Watch the trust-rebuilding space. The historical pattern for trust recovery is not persuasion campaigns or corporate “rebuilding trust” initiatives. It is structural reform forced by crisis. The FDIC did not restore banking trust by asking people to believe in banks again. It created deposit insurance — a systemic guarantee that made individual trust unnecessary. The Social Security Act did not ask workers to trust their employers for retirement. It created a structural alternative. The Wagner Act did not ask companies to negotiate with workers. It required them to. In every case, the mechanism was the same: crisis forced structural reform that made trust systemic rather than personal. In the three years between 1933 and 1938, the United States built an entirely new institutional framework — and union membership nearly tripled from 3 million to approximately 9 million.
If the Fourth Turning pattern holds, we are approaching the equivalent moment. The current trust collapse is not the end state. It is the necessary precondition for institutional rebuilding. Every Fourth Turning in the historical record ends with new institutions that command broad trust — but those institutions look nothing like the ones they replace. The FDIC did not save the old banking system. Social Security did not restore employer paternalism. Whatever emerges from this crisis — AI governance frameworks, new labor protections for tech workers, digital infrastructure guarantees — will similarly be new.
The equivalent for the current crisis will be technical infrastructure: AI transparency and accountability systems, algorithmic auditing tools, new regulatory compliance frameworks, privacy-preserving architectures, identity and trust verification systems. The developers who build that infrastructure — the structural guarantees that make trust verifiable rather than assumed — will be doing the work of institutional reconstruction. If the Strauss-Howe timeline is approximately right, the climax that forces that construction is three to six years away, and the post-crisis High that follows could run through the 2040s and 2050s. That is not a short runway. It is the kind of generational building opportunity that created thirty-year careers in the postwar era.
No one is coming to manage this transition for you. That is not a motivational statement. It is the most consistent finding across 150 years of workforce disruption evidence, confirmed by every major trust survey with decades of longitudinal data, visible in the specific institutional conditions of the software development profession today, and consistent with the cyclical pattern that places us in the gap between institutional death and institutional birth. The cavalry is not late. There is no cavalry.
But the historical pattern says something else too: every time the cavalry fails to arrive, the people who were waiting for it eventually become the builders of whatever comes next. The question is whether you are still waiting, or already building.