Files
WebinarNotes/raw/sources/sebastian interview - conclusions and insights.md

151 lines
15 KiB
Markdown
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
#ai #interview #development #webinar
# Sebastian Interview — Conclusions & Insights
**Participants:** Sebastian (founder of *Virtido*, a ~11-year-old software outsourcing/engineering company) and Eugene (the interviewer — a computer-vision & embedded/firmware developer, content creator, and builder of his own AI "harness").
**Length:** ~56 minutes · **Topic:** How AI is reshaping software engineering — teams, tooling, careers, enterprise reality, and business development.
**Context:** Appears to be a pre-webinar conversation. Casual, candid, two practitioners comparing notes.
> **Sourcing note:** Built from an automatically generated (Whisper large-v3) and speaker-diarized transcript. Quotes are lightly cleaned for readability and obvious speech-to-text errors are corrected (e.g. "Claude Code," "COBOL," "RAG," "Virtido"). A few rapid-crosstalk moments may be mis-attributed.
---
## Executive Summary
The debate is over *whether* AI can write software — it can. The whole game is now *how you use it*. That single shift cascades into everything else in the interview:
- **Teams shrink and roles merge.** The 8-person scrum team collapses to 23 people who share coordination and agent-wrangling.
- **Tooling becomes personal ("bring your own harness") — except where it can't.** Individual developers get maximum leverage from custom harnesses on top of Claude Code; regulated enterprises can't allow that, which creates a real business opportunity.
- **The value of pure coding skill collapses; the value of judgment, ownership, and relationships rises.** Claude "levels" a 20-year veteran and a fresh grad to similar output — so the differentiator moves to understanding problems and, above all, human connection.
- **The durable human skills are product ownership and real-world networking.** These are the two things AI doesn't commoditize.
---
## Who's Who
| | **Sebastian** (SPEAKER_01) | **Eugene** (SPEAKER_00) |
| -------------------- | --------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------- |
| **Role** | Founder/owner of *Virtido*, a software outsourcing company | Hands-on developer & content creator; the interviewer |
| **Vantage point** | Running a services business, client/compliance/sales lens | Building tooling, embedded/firmware & computer vision, individual-leverage lens |
| **Signature stance** | "Bring-your-own-harness won't survive enterprise; connections win business" | "Build your own harness; get closer to the client's real problem" |
| **Notable work** | 11 yrs in business, 1015 engineers on big teams, enterprise clients | Custom Claude-Code harness, projects like *insin* (apt-get-style edge updater) & *Keller* |
---
## Key Themes & Insights
### 1. The question changed: not "can AI code?" but "how do you use it?"
The cost of writing code trends toward zero, so the leverage moves entirely to *how* AI is directed.
> "It's not a question *if* AI can write software anymore — it's just a question of *how* you use it."
### 2. Team structure collapses from ~8 people to 23
No more scrum master + PM + requirements engineer + a big dev team. Instead: one coordination/ownership role plus one or two people managing coding agents, sharing responsibilities.
> "Teams will be two or three people and they will share these roles… one person with a coordination role and one or two who manage the coding agents. That's it."
### 3. "Bring your own harness" — but it hits a wall in the enterprise
Eugene's thesis (echoed by many practitioners he cites): every developer should build a personal *harness* on top of Claude Code because knowing every detail makes it far more effective. He demoed his own: a Telegram-like UI with one agent per project, inter-agent messaging, per-agent memory, and a "done thinking" signal so you don't babysit the console. Tools referenced: **Conductor** (isolates a git worktree per chat, automates PRs/merges), plus "open clock, Hermes, paperclip."
Sebastian's crucial counterpoint: this **does not scale to large or regulated organizations**. Compliance and liability make ad-hoc, per-developer setups impossible.
> "From a compliance perspective I can guarantee you that bring-your-own-harness will not be the way forward… it has to be a company-managed resource. We cannot let everyone pick their own setup."
**→ Business opportunity identified:** *scalable, manageable, company-standard harnesses for larger engineering teams.* Sebastian explicitly frames this as "the interesting market."
### 4. Enterprise reality is far more locked-down than the indie world
Sebastian's biggest clients: engineers can't use their own laptops — only a centrally managed VM, **zero ability to install their own tools.** Reference points: a Roche SAP transformation ran ~1,200 engineers for years; banks first banned AI outright and are now cautiously adopting it "because it's just so good."
### 5. Seniors are now *more* valuable; juniors are being squeezed out
Counter-intuitively, AI has raised demand for seniors and made juniors "completely irrelevant" in the market, even though a junior + Claude could in theory produce the same output.
- **Why seniors win:** 20 years of experience = knowing *where things typically go wrong*, so you don't let the AI make those mistakes. "AI does the same mistakes humans do because it's trained on our mistakes."
- **The junior risk (a security argument):** the habit of clicking "yes… yes… allow for all future" is how "API keys are leaked, databases get dumped or deleted." A junior can't evaluate a 250-line bash script; a senior at least *could*.
> "Give a junior fresh out of university access to this almighty Claude and then access to the codebase — they will [wreck] it in two days."
### 6. Ownership & product thinking is the durable human skill
The recurring "profile picture" story: an engineer implemented "change your photo," ticked every acceptance criterion, but shipped it *ugly* (visible in the corner) because they never looked at the actual result. That's the absence of **ownership**.
- Ownership = putting yourself in the user's/customer's shoes and understanding what they'll expect — a mindset/personality trait, not a task list.
- Reframe the work: **stop thinking "what needs to be done" (tickets); start thinking "what problem needs to be solved."**
> "No one ever needed a programmer… people have problems that you are solving."
- AI forces the issue: if you don't understand what to build, "you will simply not be an engineer anymore" — or you get closer to the product and the software gets *better* (product-wise, if not always technically).
### 7. The planning/waterfall debate: coordination overhead can exceed the work
A colleague (Daniel) floated returning to waterfall. Sebastian disagrees — with AI, "we don't have time for planning." Concrete evidence:
- On a small two-person project (each ~1 day/week), Sebastian is **faster alone**: the effort to synchronize who-does-what and how-it-integrates exceeds the speed of just prompting it himself.
- **Printer anecdote:** a new printer had to work in 2 days via an unfamiliar direct-printing protocol in Java (which he'd never written). Explaining it to a remote engineer wouldn't have made the deadline; he went on-site and Claude Code solved it in ~30 minutes. His edge wasn't coding — it was *knowing how to instruct and verify* the result.
### 8. Connections are everything — the standout insight
Eugene's core unsolved problem (stuck ~6 months): how to build a professional network. The framing that lands:
- Claude **levels pure programming skill.** 20 years of experience vs. a fresh grad, same Claude subscription → similar output. The *only* real differentiator becomes the ability to communicate with clients and understand their problems.
- **Real, in-person connection is the one thing AI can't commoditize** — and it gets more valuable as AI floods everything else. Soon you won't be able to tell bots from humans on LinkedIn, or even on a phone call. "In 10 years… zero" ability to tell.
- **What actually works for winning business:** showing up in person 24 days/week (business lunches, networking events, conferences, open days). **What doesn't:** sales agencies, cold calling, email marketing, LinkedIn campaigns, content, SEO — "Big zero." Every real long-term customer came through personal network.
> "The only way to get a real connection is to stand face to face in the same room with someone, shake their hand, and have a conversation."
- **Mechanics of a connection:** it forms not on the first meeting but when you meet the *same* person in *different* circumstances → recognition value → trust → referrals. ("Oh, there's this guy, Eugene — I met him a couple of times, he's real, I trust him.")
- **Be memorable in your humanness:** lead with something human (renovating a house, two kids, a cat, a dog) — not "I run a software company," which everyone forgets. Because everyone now uses the same AI tools, *everything online looks identical*; humanity is the differentiator. (Virtido's playful `humans.verti.com` / "human badge on LinkedIn" is a riff on exactly this.)
### 9. Decouple identity from profession (the philosophical turn)
Older generations tie identity to their job ("I *am* a doctor"). As many professions collapse into "prompt the AI," they'll lose obvious societal value. The advice: learn to separate *who you are* from *what you do professionally*, or "we will feel worthless in a couple of years."
> "Fundamentally I'm Eugene — I'm not a programmer. But I need to work on that."
### 10. AI's value extends well beyond programmers
Non-technical, high-leverage uses discussed: profile ~2,000 LinkedIn/HR contacts into Obsidian, build RAG/search over your own collected information, spot who's hiring and pitch as a contractor. "You can put your whole life into a RAG." You don't need to be a programmer to benefit.
### 11. Two smaller-but-sharp side points
- **Open source will grow because of AI.** When writing code costs ~nothing and most code isn't high-value, people will give it away. (Eugene's more cynical read: OSS is largely marketing; every OSS dev cites *The Cathedral and the Bazaar*.)
- **Legacy/hobby niches persist.** A few holdouts survive (e.g., COBOL engineers in banks — no training data for AI), plus people who code for the love of it "like driving an old-timer car" — but not where time, quality, and money matter.
- **On outsourcing economics (opening exchange):** great engineers exist everywhere (India has as many as Ukraine/Russia); the problem is that a *low-cost expectation* selects for bad code. Pay $10k/month and you get excellent people; an $800/month hire won't beat a good, well-paid engineer.
---
## Points of Tension / Disagreement
- **Personal vs. company-managed harness:** Eugene champions bring-your-own; Sebastian says compliance kills it at scale (and that gap *is* the business).
- **Planning:** Daniel (absent) wants waterfall back; Sebastian says coordination overhead now outweighs the work itself.
- **Open source motivation:** Eugene sees it mostly as marketing/"gambling for questionable results"; Sebastian expects *more* OSS as code becomes nearly free to produce.
- **Networking channels:** Eugene is invested in LinkedIn/articles; Sebastian bluntly calls online outreach a "waste of time" versus in-person.
---
## Standout Conclusions
1. **Move up the value chain — from writing code to owning outcomes.** The engineer's job is shifting from "produce the solution" to "understand and frame the problem, then direct and verify the AI."
2. **Seniority = risk reduction.** The senior's real product is judgment: preventing expensive mistakes and catching dangerous agent actions.
3. **The enterprise-harness market is wide open.** Someone will make money providing compliant, centrally managed AI tooling for large regulated teams.
4. **Relationships are the last non-commoditized asset.** When skill and online presence are equalized by AI, in-person trust becomes the scarce, decisive edge.
5. **Protect your identity.** Tie your sense of self to *who you are*, not a job title that AI may hollow out.
---
## Memorable Quotes
- "It's not a question *if* AI can write software anymore — it's just a question of *how* you use it."
- "No one ever needed a programmer… people have problems that you are solving."
- "Give a junior… access to this almighty Claude and then access to the codebase — they will [wreck] it in two days."
- "The only way to get a real connection is to stand face to face in the same room with someone, shake their hand, and have a conversation."
- "We tried sales agencies, cold calling, email marketing, LinkedIn campaigns, content, SEO. Zero. Big zero… Every real long-term customer has come through [personal network]."
- "We need to learn to decouple our identity from the thing we do professionally."
---
## Actionable Takeaways
**For Eugene (his stated problem — building a network):**
- Reallocate time from LinkedIn posts/articles to **recurring in-person events** (aim for the same circles repeatedly to build recognition value).
- Prepare a **memorable, human self-introduction** that isn't your job description.
- Treat the network as the moat, not the code — the coding is already commoditized.
**For engineers generally:**
- Practice **product ownership**: always look at the actual result a user sees, not just the acceptance criteria.
- Shift vocabulary from "tasks/tickets" to "problems and desired outcomes."
- Build (or learn) a **harness** for personal leverage — but expect a company-standard one at any regulated employer.
- **Read what you approve.** Don't reflexively "allow all"; that habit is a security incident waiting to happen.
**For businesses / founders:**
- The **compliant enterprise-harness** space is an underserved market.
- Assume digital outreach converges to noise; **invest in face-to-face relationship-building** as the durable channel.
---
## Open Questions Raised (Unresolved in the Interview)
- How does AI actually transform *huge* enterprise programs (the ~1,200-engineer, multi-year kind)? Sebastian is candid that he doesn't know.
- How do you unify wildly different personal workflows into one company process? (Eugene and his collaborator spent a month and couldn't.)
- Concretely, how does an individual engineer *build* a connections network from a standing start? (Sebastian offers principles; the step-by-step remains open.)