Skip to content

Tech Principles — An Epistemically Honest Defense

Open source, hardware freedom, privacy, accessibility, and equity in technology.

A commitment to transparency and accountability: this defense cites empirical research, real cases, and acknowledges gaps, limitations, and genuine trade-offs. Where evidence is mixed or inconclusive, that is stated clearly — consistent with the Epistemic Honesty directive.

Part 1 · Open source software & security

The research picture (more complex than either side claims)

The 2025 Black Duck Open Source Security and Risk Analysis (OSSRA) report, analyzing 947 commercial codebases, found the average number of vulnerabilities per codebase increased 107% in 2024. This requires careful interpretation:

  1. The complication: the surge correlates with codebase size increases (74% larger) and the rise of AI coding assistants. More code = more potential vulnerabilities, not necessarily worse code quality.
  2. The key finding: the proportion of codebases containing at least one vulnerability stayed "mostly steady year-over-year, from 86% in 2024 to 87% in 2025" — the problem isn't that open source became more vulnerable, but that applications became more complex.

Disclosure & patch management (where open source genuinely has an advantage): an ACM empirical study of 8,073 open source projects (2017–2023) found that while 80% of practitioners support Coordinated Vulnerability Disclosure (CVD), only 55% of vulnerabilities actually conform to CVD in practice — public disclosure happens before fixes are ready 42% of the time, creating temporary exposure windows.

Compare to proprietary systems: the closed-source equivalent problem is opacity. The Equifax breach (147 million people) resulted from failure to patch a known vulnerability in proprietary software. WannaCry exploited proprietary Windows vulnerabilities. Neither model is automatically secure:

  • Proprietary: vulnerabilities can remain hidden indefinitely; disclosure is controlled by vendor schedules.
  • Open source: vulnerabilities eventually become visible; disclosure is community-driven but sometimes chaotic.

Honest assessment

For open source: ✓ transparency enables faster detection (many eyes); ✓ decentralized patching means anyone can fix critical bugs; ✗ unmaintained dependencies create long-tail risk (90% of codebases had components outdated >4 years); ✗ coordination failures mean patches are sometimes disclosed before fixes exist.

For proprietary: ✓ dedicated security teams with accountability; ✗ "security through obscurity" doesn't work; ✗ end-of-life abandonment is absolute — no community can patch.

The principle: open source's structural advantage isn't perfection — it's that it distributes the ability to audit and fix. For critical infrastructure, that distribution of power matters more than any single vendor's competence.

Part 2 · Right to repair & hardware transparency

Documented economic harm

PIRG research found lack of repair access costs US farmers $4.2 billion per year in downtime and higher repair costs.

In January 2025, the FTC sued John Deere, alleging that limiting diagnostic software (ServiceADVISOR) to company-affiliated dealers is an unfair, illegal practice that "boosted Deere's multi-billion-dollar profits." Specifics:

  • John Deere and CNH Industrial control nearly 90% of the US market for large tractors and combines.
  • The farmer-available version of ServiceADVISOR is "neither robust nor effective."
  • The FTC complaint describes Deere's practices as "intentional and strategic, approved by company executives."

Settlement: in 2026 John Deere settled a class action for $99 million and committed to 10-year diagnostic-tool access, while maintaining "no finding of wrongdoing." The FTC case remains pending.

Legislative momentum: Colorado's farm-equipment Right to Repair law took effect January 1, 2024; California and Minnesota enacted laws on electronics repair.

What this means

The Deere case reveals the structural problem: a manufacturer uses software locks to enforce a repair monopoly — not about quality, but about controlling aftermarket revenue. Farmers cannot audit or modify their own equipment, diagnose problems independently, or choose cheaper repairs. The manufacturer frames it as "emissions compliance" and "security," but the effect is monopoly pricing. If hardware were transparent (source available, firmware modifiable), an owner could diagnose, repair, modify for conditions, and extend functional life. A US District Judge ruled in 2025 that Deere must face claims it violated Colorado's law — confirming genuine legal violation, not just preference.

Honest complication

Manufacturers raise concerns that are sometimes legitimate: emissions systems (owners can modify to pollute more), safety hazards, and genuine supply-chain complexity. The honest counter: these don't justify total information lock-down. Elizabeth Warren noted Deere omits repair-rights and pollution-control information from manuals, potentially violating the Clean Air Act. The principle: you can enforce standards and enable repair — one doesn't require the other.

Part 3 · Privacy & behavioral manipulation

What large-scale data collection actually does

A 2023 Pew study: 73% of Americans believe they have little/no control over what companies do with their data, 67% have no idea what businesses do with it, and 81% are very or somewhat concerned about its use.

This is backed by economic research: a 2025 study by Acemoglu, Makhdoumi, Malekian, and Ozdaglar (American Economic Review) found AI tools let platforms learn product "glossiness" (attributes that make low-quality products appear attractive) and engage in behavioral manipulation; when glossiness is long-lived, manipulation reduces user welfare. Concretely: platforms identify what makes you click on low-value content, optimize feeds to show more of it regardless of benefit to you, and profit from engagement while you lose time and money — and you cannot audit or resist this because the algorithms are proprietary and the data asymmetric.

Why this isn't purely about secrecy

The privacy argument isn't "I have nothing to hide" — it's structural:

  1. Asymmetric power: you cannot audit or resist invisible optimization.
  2. Chilling effects: when you know you're watched, you self-censor.
  3. Behavioral modification at scale: these systems don't inform you — they nudge you toward actions that benefit the platform.
  4. Economic extraction: your attention and behavior are converted directly into corporate revenue.

State data-privacy laws have proliferated (Texas and Oregon July 2024; Montana October 2024; Delaware, Iowa, New Jersey, New Hampshire, Tennessee in 2025), reflecting growing recognition that current practices are unsustainable.

Transparency would look like: open source platforms (code auditable by anyone), clear data policies (what's actually collected), user control (opt out of behavioral tracking), and algorithmic transparency. The principle isn't to prevent all data use — it's to prevent invisible, non-consensual behavioral modification.

Part 4 · Accessibility & digital equity

Scale of the problem

An estimated 16% of the global population — ~1.3 billion people in 2024 — has a significant disability. In the US, 26% of adults (61 million people) have disabilities. This is not a niche issue.

Economic impact of inaccessibility

Over 96% of the top one million web pages had accessibility issues in 2023. Business impact:

  • Average cart abandonment: 69% on inaccessible sites vs. 23% on accessible sites.
  • ROI of accessibility: $100 return for every $1 invested (Forrester, 2024).
  • People with disabilities control ~$490 billion in disposable income.

This is not charity — it's market opportunity.

Employment disparities (the equity piece)

People with disabilities earn a median 66 cents per dollar earned by people without a disability. Their unemployment rate is 7.2% vs. 3.5% — more than double. Inaccessible job applications, training materials, and e-commerce platforms systematically exclude them: when job sites don't work with screen readers, applications require mouse-only interaction, or training has no captions, the exclusion is architectural, not incidental.

In 2022, US courts saw over 4,060 web accessibility cases (+76% from 2018). In H1 2025 alone, more than 2,000 ADA website lawsuits were filed (+37% YoY). The average web page contains 297 accessibility issues failing WCAG criteria, and a web accessibility lawsuit is estimated to cost $100,000.

Technical reality

Accessibility failures aren't mysterious: low-contrast text affected 79.1% of homepages (easily fixable); missing alt text affects 55.5% of pages — 44% of those involve linked images, completely breaking screen-reader navigation (easily fixable). These are simple design oversights that create total exclusion.

What "commitment to accessibility" means

Designing with accessibility from the start (not retrofitting); testing with real disabled users (not just compliance checkers); understanding captions help everyone (noisy environments, language learners); recognizing accessibility as a feature serving a large, economically significant population. When 26% of a potential user base faces systematic barriers, fixing them is urgent, not optional.

Part 5 · Digital equity & long-term consequences

Current disparities (documented)

A 2024 Pew study found only 57% of households earning under $30,000 subscribed to broadband, vs. 76% in the $30,000–$69,999 bracket. A 2024 Pew Charitable Trusts report indicates 43% of adults earning under $30,000 lack broadband access, and nearly half of households under $50,000 struggle to afford internet.

The "homework gap"

COVID-19 lockdowns exposed the disparity: some kids had school-issued Chromebooks but no wifi; others had no device. In 2024 the US Department of Education's National Educational Technology Plan identified three divides: digital access, digital design, and digital use. The critical finding: handing out devices isn't enough — the "digital use divide" is about how technology is used (active deep learning vs. passive assignment completion). Students with good access plus good pedagogy develop skills faster; the gap compounds over time (lack of access in 2nd grade limits 8th-grade options, which limits college options).

Why this connects to open source

If educational software is proprietary, schools buy licenses they cannot modify, teachers can't see how algorithms make recommendations, students learn only by consuming, and product abandonment means total loss of access. If software is open, schools adapt tools to context, communities build their own capacity, students learn by building and modifying, and the tool remains available indefinitely.

Long-term impact

Disability rates rise sharply with age (16% of adults 18–44, 29% for 45–64, nearly 50% for 65+). Black adults have a 31% disability rate and Hispanic adults 30%, vs. 24% among white adults — reflecting systemic health disparities. So digital-equity barriers disproportionately hit communities already facing disadvantage; early exclusion creates lasting consequences; and accessible, open tools become increasingly critical infrastructure.

Part 6 · The coherence argument & where it breaks down

Why these principles reinforce each other

  1. Open source + transparency enables auditing for both security and privacy violations.
  2. Hardware freedom + transparency enables privacy (remove surveillance features) and equity (communities modify tools).
  3. Privacy protection enables equity (prevents profiling and manipulation of vulnerable populations).
  4. Accessibility ensures equity (removes architectural barriers).

Where tensions actually exist (honest limitations)

Maintenance burden. Open source is largely maintained by a handful of volunteers while 99% of the world's software relies on it. More transparency means more people can audit but also more bug reports volunteers must handle; rootkit-able hardware means more security surface; accessibility requirements increase development cost. Honest answer: real trade-offs — the solution is to recognize costs and build sustainable funding models.

Security vs. convenience. Root access enables modification but also lets users break their own systems; perfect transparency means more attack-surface detail is public; accessibility features can conflict with performance optimization. Honest answer: the question is who controls the trade-off — users or corporations — and the commitment is to make it visible and controllable.

Scale and complexity. Full transparency of a modern CPU's microarchitecture is complex enough to be nearly useless to most people; full firmware access means most people will brick their devices. Honest answer: transparency doesn't require everyone to understand everything — it requires that experts can audit and that intentional obfuscation is prevented.

Part 7 · What would evidence of success look like?

Open source: 80%+ of critical infrastructure maintained by more than 2 communities; coordinated disclosure before public release in 80%+ of cases; community patches for abandoned software within 30 days. Current state: mixed.

Hardware freedom: diagnostic/repair info available for all major products; root access without voiding warranty for non-safety-critical systems; owner-modifiable firmware. Current state: improving (right-to-repair laws passing) but most hardware still locked down.

Privacy: data collection limited to stated purposes; behavioral tracking transparent and user-controlled; data brokers regulated like financial institutions. Current state: minimal; collection accelerating.

Accessibility: 95%+ of major websites WCAG compliant; legal cases declining as conformance becomes standard; accessible design treated as required baseline. Current state: ~5% of websites fully WCAG compliant; 2,000+ lawsuits/year.

Equity: broadband access >99% (like electricity); open tools funded as public infrastructure; digital literacy accessible to all demographics. Current state: 43% of low-income households lack broadband.

Conclusion

The "fanaticism" isn't about purity or ideology. It's about recognizing structural problems that concentrate power (a few companies control infrastructure billions depend on, with no transparency), exclude systematically (current designs exclude disabled, low-income, and entire communities), enable manipulation (asymmetric data and closed algorithms allow behavioral modification at scale), and prevent repair (manufacturers force obsolescence by locking access).

These are engineering choices, not natural constraints. The evidence shows open source can be as secure as proprietary when well-maintained; right to repair creates measurable benefit ($4.2B/year for farmers); behavioral-manipulation research confirms asymmetric data creates real welfare loss; 1.3 billion disabled people are systematically excluded; and digital gaps compound into long-term disadvantage. Where evidence is mixed, that's stated. Where trade-offs exist, those are acknowledged. The core claim stands: the current system distributes power and opportunity unequally in ways that are measurable and addressable.

Sources & methodological notes

Research standards used: peer-reviewed studies (ACM, American Economic Review, 2024–2025 where possible); government data (FTC, Department of Education, CDC); industry reports (Black Duck, Forrester, WebAIM); actual litigation (John Deere case, accessibility lawsuits); quantified claims only.

Where evidence is weak or contested: proprietary-vs-open-source security comparison (both sides overstate advantages); accessibility-software effectiveness (overlays not reducing litigation; real compliance improving slowly); maintenance sustainability (volunteer burnout; no solved funding model); hardware complexity (full transparency technically impossible for modern systems).

What's not in this document: speculation about future technology; anecdotes or single examples (uses aggregated data); attacks on people with different views; claims beyond what research supports.

The goal: defend principles with evidence, acknowledge limitations honestly, and recognize that technology choices have real consequences for real people.