Summary of The Pragmatic Engineer AMA

Overview of The Pragmatic Engineer AMA

In this AMA, Gergely Orosz answers listener questions about his career transition from Uber to writing and research, the impact of AI on software engineering, hiring trends, big tech culture, and how he runs The Pragmatic Engineer as a business. The recurring theme is pragmatic adaptation: AI is changing workflows, but fundamentals like product sense, craftsmanship, and business judgment still matter. He also shares candid thoughts on company culture, career strategy, education, and what he’s learned from building a media business around engineering content.

Why Gergely Left Uber and Started Writing

Gergely explains that his move away from a full-time engineering/management role happened because several factors lined up:

• Uber’s COVID-era layoffs and business decline made the work feel less motivating.
• He had financial runway from Uber equity, giving him room to take a risk.
• He was tired of middle-management tradeoffs and wanted more ownership.
• He had already started writing a book, blogging, and making videos on the side.
• A startup idea didn’t feel compelling enough to justify a 10-year commitment.

Ultimately, he realized that writing and sharing engineering knowledge was not just a side activity—it was something he genuinely wanted to do. He tested the newsletter model, it took off, and the rest followed.

AI’s Impact on Software Development and SDLC

Gergely sees AI changing the software development lifecycle, but not in a fully standardized or universally “AI-native” way yet.

What he’s seeing in practice

• Big companies are using AI to build better internal tooling, especially coding agents connected to company systems.
• Anthropic is the clearest example of a company operating in a fluid, prototype-heavy, AI-native way.
• Most large companies are not fully “retrofit” around AI; instead, they’re adding AI infrastructure and workflows incrementally.

His core view

AI speeds up development, but:

• it doesn’t replace planning entirely,
• it doesn’t remove the need for product and reliability thinking,
• and it can’t justify breaking mature business processes just for speed.

He emphasizes that AI should be treated as another tool in the stack—not a religion.

Hiring in the AI Era

A major topic in the AMA is how AI is reshaping hiring.

What companies are optimizing for now

Gergely says hiring is shifting toward candidates who can:

• reason through AI-generated output,
• detect when the model is wrong,
• do their own research,
• and make informed architecture decisions.

How hiring may evolve

He believes:

• take-home tasks will increasingly assume AI usage,
• but final decisions will move more toward in-person or live evaluation,
• and traditional whiteboard/interview loops may regain importance because they’re harder to outsource to AI.

What this means for candidates

Hiring will likely feel:

• more subjective,
• less standardized,
• and more frustrating.

His advice: expect more friction and invest more effort into preparation.

Which Engineers Are Thriving

Gergely says the most in-demand engineers today tend to be:

• product-minded,
• working at startups or recognizable tech companies,
• comfortable building with AI,
• and able to help decide on architecture and model strategy.

In particular, people who have built real AI-powered systems are now highly sought after. Companies want engineers who can help answer practical questions like:

• Should we use RAG?
• Fine-tuning or off-the-shelf models?
• On-prem or hosted inference?
• How do we manage inference costs?

He contrasts this with engineers who have little AI exposure or come from less “modern” company pedigrees, who may find it harder to make the jump.

Big Tech, Culture, and AI Adoption

Gergely gives blunt takes on how major tech companies are handling AI:

Google

• Most aggressive overall in AI adoption.
• Gives teams room to build internally.
• Gemini is the only major consumer AI product meaningfully competing with ChatGPT.

Meta

• He thinks leadership, especially Zuckerberg, is operating in wartime mode.
• Morale appears to be declining because employees don’t always understand the urgency.
• Some engineers are being shifted into work they don’t find meaningful.

Microsoft

• He sees too much politics and organizational confusion.
• AI strategy appears fragmented across groups.

Apple

• He thinks Apple is largely quiet and possibly underwhelming on AI.
• But their focus on devices and hardware may still serve them well.

Amazon

• A strong example of how hard it is to retrofit innovation into a large org.
• They are trying hard, but AI adoption feels slow and bureaucratic.

The companies he thinks are doing better

Smaller public companies and scaled product companies like:

• Uber,
• Ramp,
• Intercom,
• Block

These companies can adopt AI pragmatically without needing to own the whole stack.

AI-Native Companies vs. Pragmatic Adoption

Gergely is skeptical of “AI-native” as a blanket strategy.

His view

• Anthropic is a special case, not a template most companies can copy.
• Many companies are better off using AI where it naturally solves a real problem.
• Forcing AI adoption just because it’s trendy is often a mistake.

Better framing

Instead of asking, “Are we AI-native?” ask:

• Where does AI help this specific problem?
• Where does it improve speed, debugging, or support?
• Where should we ignore it?

He thinks companies should treat AI like any other tool: useful when it fits, discard it when it doesn’t.

Code Quality, Tech Debt, and Speed

A key discussion is whether AI encourages bad code and weaker engineering quality.

Gergely’s answer

He argues the speed-vs-quality debate is too simplistic.

• Early-stage companies may benefit from speed and prototypes, even with tech debt.
• Growth-stage companies often accept hacks to scale faster.
• Mature companies should care much more about reliability and maintainability.

He uses Uber’s old architecture as an example of a terrible but effective workaround that unblocked the business. His broader point: tech debt can be valuable early, but quality matters more as the product matures.

He also notes that AI can help not just build faster, but refactor faster—which reduces the excuse to avoid cleanup.

Engineering Managers in the AI Era

Gergely is not convinced there is one correct model for EMs.

Two styles

• Non-coding EMs: focus more on people, systems, and organizational issues.
• Coding EMs: bring more technical depth and can guide architecture better.

Trend he sees

The industry is moving toward expecting EMs to be more technical. That may reduce the amount of people-centered management support engineers get for a while.

He expects the pendulum may swing back later, but for now technical credibility is rising in importance.

What Actually Proves AI Productivity Gains

Gergely is skeptical of vanity metrics like:

• token usage,
• code volume,
• PR count,
• or leaderboard-style AI adoption tracking.

What matters instead

He suggests AI productivity should show up as either:

• incremental revenue,
• or real cost savings.

If AI merely makes it easier to do things that would have happened anyway, that’s not necessarily meaningful productivity gain.

He also notes that many of the strongest AI business cases today are in:

• AI products themselves,
• support and operations,
• incident response,
• and infrastructure automation.

Education, Degrees, and Hiring Prestige

Gergely believes degrees and university prestige matter more now than during the bootcamp boom.

Why

• The market is more competitive.
• Bootcamps no longer reliably lead to jobs.
• Prestigious universities still matter for access to top employers.
• Degrees can also matter for visas and immigration.

His advice: if you’re a student, try to get internships, build impressive projects, and aim for strong signals of competence.

Practical Career Advice for Different Situations

He gives several pieces of actionable advice:

• Junior engineers: get any real job first; then excel and build network/experience.
• Engineers in slow companies: try to start AI experiments internally.
• Engineers feeling stuck: move closer to teams or companies where modern tooling and AI are actively used.
• Students: build side projects, especially if you want to stand out.
• People in boring or rigid environments: the fastest path to relevance is hands-on experience, not another degree alone.

Gergely’s Own Workflow and AI Usage

He says he mostly uses AI for:

• research,
• coding his own internal business tools,
• and experimentation with multiple tools.

Tools he mentions

• Codex
• Claude Code
• Cursor
• Factory AI
• GPT-5.5

What he does not use AI for

He deliberately avoids using AI for his writing because:

• he enjoys the thinking process,
• the output doesn’t sound like him,
• and writing is where he develops ideas.

He worries more about weakening hands-on coding fluency than writing ability, and he’s okay with that tradeoff.

The Pragmatic Engineer as a Business

Gergely shares some business context:

• The newsletter has grown to over 10,000 paying subscribers.
• He also has podcast sponsors.
• He doesn’t like sharing exact revenue too often because it invites coaching/mentoring requests.
• He likes being in control of the business and setting the pace himself.

Future plans

He wants to:

• make Pragmatic Summit a more regular event,
• possibly host one in Europe,
• and slowly grow the team.

He’s not trying to expand aggressively just for the sake of scaling.

Journalism, Risk, and Articles That Nearly Got Him in Trouble

Gergely talks about the limits of his writing and reporting.

One article he never published

He nearly published a highly negative piece about Bunq in the Netherlands, but decided against it because it felt one-sided and didn’t help anyone.

One article that caused real stress

He wrote about layoffs and mistreatment at a company called Poland/Polina (likely a transcript error; the context suggests a travel/event company). The company threatened legal action, and he had to heavily edit for libel concerns.

This taught him that investigative journalism is not the path he wants to pursue.

Favorite Books and Products

Favorite technical books

• The Philosophy of Software Design
• Tidy First? by Kent Beck

Favorite products

• Granola for meeting notes
• Perplexity for search and deep research

He likes products where AI creates a genuinely new experience, not just a gimmick.

Long-Term Beliefs About Software Engineering

Gergely ends with a broad prediction: the industry will still need true professionals.

What won’t change

• people who understand tools and tradeoffs,
• engineers who care about correctness and craft,
• and professionals who can choose the right tool for the job.

He believes AI may actually filter out people who never cared deeply about software in the first place, while increasing demand for those who do.

Key Takeaways

• AI is real and useful, but it should be applied pragmatically, not dogmatically.
• Hiring is becoming more subjective and more AI-aware, with live evaluation regaining importance.
• The best engineers are those who combine AI fluency with product judgment and technical reasoning.
• Big tech adoption of AI is uneven; smaller product companies may be better positioned to adopt quickly and sensibly.
• Quality still matters, but the right balance between speed and craft depends on company stage.
• The strongest career move today is to get hands-on with modern tools inside a real product environment.

Sponsor Mention

The episode is sponsored by Antithesis, which Gergely highlights as a tool for verifying system correctness through hostile simulation, bug probability analysis, log exploration, and replayable debugging.