Software Engineering Career

Summary

A recurring undercurrent in Mr. Nayak’s catalog: what happens to software engineering as a profession when AI can generate most of the code? Sources range from philosophical (is it still a lifetime career?) to personal (laid off by Atlassian) to practical (reading code, interview prep tools, the vibe coding debate). The picture is nuanced - AI doesn’t eliminate the engineer, but it reshapes what “being a good engineer” means. W23 adds three new angles: the Dead Economy Theory frames AI investment as a structural labor displacement play; domain expertise as the non-fungible moat (Horsting: the scarce resource is whether you can verify AI output, and that requires owning a domain model); and the “do the hardest thing” principle as a career strategy response - choosing ambitious problems with few competitors as a hedge against commoditization. W24 adds two contrasting responses to the AI pressure. North Star (Loris Cro) is a values clarification: when AI commoditizes code production, the durable north star is maximizing user utility — not technical mastery, not DX, not abstraction beauty. The LLMs-eroding-career piece is the most emotionally direct account in the catalog: not structural analysis but a first-person record of watching accumulated expertise become replicable on demand, in the exact domain you chose to build depth in. W25 adds Paul Graham’s How to Earn a Billion Dollars (Oxford Union, June 2026): an articulation of the startup-path wealth creation mechanism (user value → exponential growth → compounding valuation) as a direct rebuttal to claims that billionaire wealth is structurally exploitative. W26 adds Charity Majors’ career signal: engineers who raise their standards in response to AI capability gains — treating it as a quality accelerator, not just a velocity accelerator — will be structurally more valuable than those who lower them.

W27/W28 adds the career-health dimension to AI coding: the replacement narrative remains mostly marketing when it ignores expert judgment, but the workflow itself can still erode boundaries through addictive prompting loops and scope inflation. The durable career response is not only “raise standards” but also preserve attention, recovery, and a clear sense of what expert work actually is.

W28 adds a small but practical career-learning note: even in an AI-assisted environment, interview and algorithm readiness still depends on fundamentals before problem grinding. The advice mirrors the broader catalog pattern: do not outsource the substrate of understanding.

W29 broadens career durability from code skill to agency, verification capacity, and health. Cohen’s affirmative path is to make domain workers the accountable managers of organization-owned agents rather than replacement targets. Park argues that mathematical understanding must be preserved even when machines can produce research results. Van Sprundel’s first-person account is the grounding counterweight: work discipline and AI-assisted throughput cannot substitute for mental health, communication, careful testing, and the capacity to finish safely.

Key Sources

W29 2026 · 11-Jul-26 → 17-Jul-26

• Don’t Go Quietly Into the AI Night — Cohen rejects headcount reduction as the primary AI strategy and proposes training existing domain experts to become accountable agent managers; the career transition is from manual task execution toward objective-setting, delegation, feedback, review, and ownership of results (thought-leadership · #software-career, #future-of-work, #agent-management, #domain-expertise, #human-centered-ai)
• Automation Without Understanding — Park argues that producing mathematical results is not the same as sustaining people who can understand, verify, and challenge them; weakening the education and research pipeline while AI automates theorem work risks destroying a strategic human capability that cannot be rebuilt on demand (paper · #software-career, #mathematics, #education, #ai-automation, #verification)
• Prioritize mental health — Ramon van Sprundel’s candid account of severe depression, repeated work failures, poor communication, incomplete/sloppy delivery, and a decision to step away from software while recovering; LLMs made multitasking easier but also removed the natural path that had forced testing, so faster completion did not solve reliability or wellbeing (opinion · #software-career, #mental-health, #developer-wellbeing, #communication, #ai-coding)

W28 2026 · 04-Jul-26 → 10-Jul-26

• Fernando Franco: 12 DSA fundamentals before LeetCode — LinkedIn post: engineers struggle with DSA when they jump straight to LeetCode; recommends one-week-at-a-time fundamentals across recursion, complexity, arrays, linked lists, hash tables, BSTs, priority queues, stacks/queues, graphs, sorting, dynamic programming, and pattern recognition; useful as a learning-path reminder amid AI-assisted shortcuts (other · #software-career, #interview-prep, #dsa, #learning, #fundamentals)
• AI coding is addictive. Engineers are paying the price — LeadDev: engineers are not simply getting time back from AI; many are working more hours, with advanced engineers most affected, and emotional drain rising across ICs and leadership; career durability now includes managing the “AI Vampire” loop through time-boxing, separating exploration/execution, and treating recovery as maintenance (opinion · #software-career, #ai-coding, #burnout, #work-habits, #developer-productivity)

W27 2026 · 27-Jun-26 → 03-Jul-26

• Give Smart People The Tools To Do Smart Things — superuserdone: replacement claims reduce professional work to visible artifacts; compilers did not replace programmers and AI should be framed as expert amplification, because the hard parts remain understanding, tradeoffs, consequences, and judgment; a career framing that validates expertise rather than treating it as luddite resistance (opinion · #software-career, #expertise, #ai-coding, #future-of-work, #human-judgment)

W26 2026 · 20-Jun-26 → 26-Jun-26

• AI demands more engineering discipline. Not less — Charity Majors: the engineers who thrive in the post-threshold AI era are those who raise standards in response to AI, not lower them; code review volume is now the bottleneck, not code production; a career signal — engineers who treat AI as a quality accelerator (not just velocity) will be structurally more valuable (opinion · #software-career, #ai-coding, #engineering-discipline, #code-review)

W25 2026 · 13-Jun-26 → 19-Jun-26

• How to Earn a Billion Dollars — Paul Graham, Oxford Union talk (June 2026): billionaires from startups earn wealth by creating things users love, not through exploitation; 21 years of YC, 6500 companies funded, 30+ billionaires; the mechanism: users love → exponential growth → compounding valuation; rebuttal to a politician’s claim that “it’s impossible to earn a billion without doing something bad”; germane to software-career as the dominant mechanism for startup-path wealth creation for engineers (thought-leadership · #software-career, #entrepreneurship, #startups, #yc, #wealth-creation)

W24 2026 · 06-Jun-26 → 12-Jun-26

• My Software North Star - Loris Cro’s priority stack: (1) useful to the end user (“software you can love”); (2) correct; (3) maintainable and efficient; memory safety, abstractions, and DX only matter in service of maximizing end-user utility; a values-first articulation as counterweight to AI-accelerated output-for-output’s-sake (opinion · #software-career, #engineering-fundamentals, #software-quality, #user-value)
• LLMs are eroding my software engineering career and I don’t know what to do - 10-year fintech engineer: employer-mandated AI adoption rendered accumulated domain expertise (PCI compliance, payment idempotency, ledger design) replicable by models on demand; erosion moved beyond domain knowledge into communication identity; not a fear of the future - a documented present experience in a field the author specifically chose for its domain-depth differentiation (opinion · #software-career, #human-ai-collaboration, #ai-coding, #domain-expertise, #career-anxiety)
• LLMs are eroding my software engineering career - Hacker News - (unreachable: 429 rate-limited) HN thread; inferred debate: documented domains (finance, payment processing) vs. tacit-knowledge domains as differential AI resistance; whether seniority provides a floor; structural displacement vs. tool adjustment framing (hn-thread · #software-career, #ai-coding, #career-anxiety)

W23 2026 · 30-May-26 → 05-Jun-26

• Domain Expertise Has Always Been the Real Moat - Horsting: the engineer’s traditional advantage was translating a domain model into code; agentic AI collapsed the translation step but not the domain model acquisition step; now the scarce resource is “can you tell whether it’s right” - and only someone who holds the domain deeply can answer; a logistics dispatcher using an agent outperforms a generalist engineer in that domain because they supply the oracle; engineers need to reckon with which domains they actually hold deeply enough to audit AI output (thought-leadership · #software-career, #ai-coding, #domain-expertise, #human-ai-collaboration)
• The Dead Economy Theory - the AI industry’s financial model requires displacing the global labor market at scale; the gentle framing (“augmentation”) is marketing for what investors are actually funding; every “does the work of ten analysts” pitch is a labor-replacement product pitch; framing that matters for software engineers assessing long-term career risk (thought-leadership · #software-career, #labor, #ai-impact, #future-of-work)
• Do the hardest thing - Justin Jackson + Jesse Hanley (Bento founder): there is little competition in hard, ambitious problems because most people avoid them; choosing the hardest path in a niche leads to durability (7 years of grinding on Bento); advice to stop restricting projects to what you can do solo and to pick challenges with real upside; a practical career strategy in an era of commoditized code production (opinion · #software-career, #entrepreneurship, #strategy, #persistence)

W22 2026 · 23-May-26 → 26-May-26

• The Revenge of The Measurers - blunt essay: May 2026 tech layoffs are here (Meta -8,000, Cloudflare first mass layoff in 16 years, 100k+ YTD); the claim “they just overhired during COVID” has been quietly dropped by VCs; argues AI is cutting the organizational middle first - the “measurers” (middle management, finance, legal, internal audit) - because their function was managing complexity that AI now handles; Peter Drucker framing: only producers and marketers create value, everyone else is cost (opinion · #layoffs, #ai-impact, #future-of-work, #white-collar-jobs)
• Amazon Web Services - Four Years and Out - personal account of being fired from AWS after 4 years on the open source strategy team; describes Amazon’s view of employees as “fungible” (replaceable); hired by a manager who called him “non-fungible” (a specialist); that manager was reorganized out; the tension: Amazon’s fulfillment-center HR model doesn’t translate to institutional knowledge roles; also cites GenAI over-focus as a secondary driver of dissatisfaction (opinion · #aws, #open-source, #big-tech, #career, #layoffs)
• Reverse centaurs and the failure of AI - Cory Doctorow: centaurs = human+machine doing more than either alone (chess grandmasters + software); reverse centaurs = machine using human for support, not the other way around; Amazon Mechanical Turk is the archetype; warehouse robots didn’t remove physically punishing tasks, they made workers pace-setters for machines; a 2021 concept that has aged well as a frame for AI-accelerated labor transformation (opinion · #human-ai-collaboration, #automation, #labor, #ai-ethics)
• Coding agents are giving everyone decision fatigue - AI makes code generation cheap but makes human judgment more expensive; work density up, not down; the new scarce resource is the ability to define “what good looks like” (opinion · #ai-coding, #decision-fatigue, #developer-productivity)
• Thoughtworks: Future of Software Engineering Retreat - industry retreat key takeaways; referenced in the olano.dev pieces; covers how code review and quality processes need to change as LLM output exceeds human review capacity (thought-leadership · #software-engineering, #ai-coding, #future-of-work)
• Responsible Work - Satisfice PDF resource on professional responsibility in software work; testing-adjacent; a reference document on what accountable engineering practice looks like (other · #software-quality, #testing, #professional-ethics)

W21 2026 · 16-May-26 → 22-May-26

• Why I Don’t Vibe Code - experienced developer’s personal case against vibe coding: cost friction (token economics feel absurd to a self-described cheapskate), experience-based calm, workflow mismatch; the piece is valuable as a genuine counterpoint from someone with deep experience, not a technophobe
• SharpSkill vs LeetCode - comparison of interview prep tools; SharpSkill targets real-stack mastery (React, Node, Spring, etc.) for non-FAANG interviews; LeetCode targets FAANG algorithm grinders; reflects how AI has shifted interview prep market toward application-layer skills
• I was laid off by Atlassian - personal account (YouTube) of an Atlassian layoff; context in broader tech industry contraction
• Reading code instead of writing code: The underestimated senior discipline - as LLMs generate more code faster, code reading becomes the critical differentiator; senior engineers must audit, understand, and navigate code they didn’t write
• Don’t Outsource the Learning - skill atrophy is subtle: thousands of small copy-paste interactions; engineers who use AI for conceptual questions retain comprehension; copy-pasters don’t

W20 2026 · 09-May-26 → 15-May-26

• Software engineering may no longer be a lifetime career - Goedecke’s structural argument: the learn-by-doing incentive structure is disrupted by AI; not because AI makes you dumber, but because the economic incentive to learn deeply through practice is being dismantled

W19 2026 · 02-May-26 → 08-May-26

• ad-si/Coding-Flashcards - flashcard-based coding practice; noted as potentially useful for interview preparation as a complement to more hands-on tools

Open Questions / Tensions

• Health before productivity systems: Van Sprundel’s account resists reducing performance problems to a better checklist, prompt, or agent workflow. Planning and communication matter, but severe depression changes the capacity to apply them; career advice must distinguish skill/process gaps from health conditions that require time and professional support.

• Agent manager as a mass career path: Cohen’s proposal preserves domain workers by moving them into delegation and review, but it assumes most people can and want to become managers of automated labor. It is unclear whether this produces broader agency or simply universalizes the oversight burden already visible as botsitting.

• Fundamentals before acceleration: The DSA post is the interview-prep version of ‘don’t outsource the learning.’ AI can explain or generate solutions, but career confidence still depends on mastering recursion, complexity, data structures, and problem patterns yourself.

• Career resilience now includes boundary-setting: Majors’ W26 answer was “more discipline”; LeadDev/Yegge adds “more recovery.” If AI makes high-output work feel continuously available, career durability depends on resisting the dopamine loop as much as on improving code review or domain judgment.

• Learn-by-doing is broken, now what? Goedecke identifies the structural problem but doesn’t fully resolve it. Harris’s “Why I Don’t Vibe Code” shows that experienced engineers may simply opt out - but that’s not a solution for juniors who need the learning that practice provides.

• Atrophy vs. augmentation: Goedecke and Osmani describe the same phenomenon from different angles - one as economic structure, the other as personal discipline. Both are right. Harris adds a third angle: for some, the economic argument (cost of tokens) is itself the deciding factor.

• Interview prep market shift: SharpSkill vs LeetCode reflects a real fork in the market - are engineers being hired to solve abstract puzzles or to master their actual stack? The answer differs sharply between FAANG and the rest of the industry.

• Layoffs and AI causation: The Atlassian layoff video is data in a broader trend. Hard to attribute individual layoffs to AI directly, but the macro signal - large tech companies shedding engineering headcount - is real and ongoing.

• Domain expertise vs. generalist T-shape: Horsting’s argument challenges the standard T-shaped engineer ideal. In a world where AI handles breadth, the human value-add is depth - but depth in a domain, not just in software. The engineer who pairs software judgment with genuine domain expertise (legal, medical, logistics) may be structurally more durable than one who pairs it with breadth across many technical domains.

• “Hardest thing” as a commoditization hedge: Jackson’s advice predates the current AI discourse, but maps onto it cleanly: if AI commoditizes easy software tasks, the protected value is in problems hard enough that few are willing to attempt them - which often correlates with high domain specificity (the Horsting connection).

• North Star as a stability anchor: Cro’s values hierarchy (user utility → correctness → maintainability) provides a stable evaluation frame when AI changes what’s easy to produce. If the output is cheap to generate, the north star becomes the standard by which you evaluate it - and that standard (does it actually serve the user?) is AI-invariant.

• The LLMs-eroding account as a Horsting stress-test: The bearblog piece directly tests Horsting’s “domain expertise as moat” thesis and finds it partially wanting: the author had domain depth (fintech), and that depth was still eroded by AI within 12 months of adoption. The caveat: they describe documented, well-structured domains (payments, PCI compliance). Horsting’s oracle claim may hold for high-tacit-knowledge domains better than for well-documented ones.

• PG’s meritocracy argument in the AI era: Graham’s Oxford Union piece argues the wealth-creation mechanism (user value → growth → wealth) is legitimate. The counter-thread in the catalog (Doctorow’s reverse centaurs, Horsting’s oracle argument, bearblog’s erosion account, McGrann’s Dead Economy) suggests the mechanism itself is changing: AI enables value creation with fewer humans in the loop, which changes the structure of wealth distribution even if the mechanism remains technically legitimate. The question isn’t whether the billionaire made something users love — it’s how many engineers were needed to make it.

• Engineering discipline as the differentiator: Majors’ piece is the affirmative career answer to the LLMs-eroding narrative: the engineers who thrive are those who raise standards, not those who resist AI. This connects to the North Star principle (Cro): when code production is cheap, the evaluation standard (does this actually serve the user?) becomes the scarce and valuable contribution.