In 2026: Fewer Engineers, More Product Managers - Daniel Marcous (April)
Host: Eyal David · Guest: Daniel Marcous · About the host
Daniel Marcous, Co-Founder and CTO of April (which builds an AI-based tax suite for American citizens) and formerly the founder of the CTO Office at Waze/Google, plans to hire fewer engineers and more product managers and designers in 2026. The reason: AI tools and vibe coding are expected to raise engineering velocity by 1.3x to 1.5x, so the next bottleneck is product managers who can't produce precise requirements fast enough. In his view, what sets people apart in this era is a builder mindset, courage and ownership, not necessarily technical ability.
In this episode
- When AI raises engineering velocity, the bottleneck moves to product managers. That's why Daniel plans to hire fewer engineers and more product managers and designers in 2026, changing the 1-to-10 ratio he used to run with.
- When working with agents, the most important thing to read is the reasoning, not the code. Almost every failure comes from missing context, and every wrong decision by the agent shows what needs to be added to the knowledge base.
- A team of agents doesn't replace people one for one: one person manages about three features in parallel, and a power user costs about 1,000 to 2,000 dollars a month, an investment with a much higher ROI.
- Product managers, designers and even CS and Finance teams can build small features and internal apps end to end, as long as the organization gives them a clear path to production with a limited blast radius instead of extinguishing their creativity.
- A builder mindset, courage and ownership matter more than a technical background. In interviews Daniel keeps piling on difficulties and looks for candidates who question the problem definition itself, and a candidate who submitted a working app for her take-home assignment got the job.
- When code is no longer a moat, a spray-and-pray strategy of many experiments in production becomes legitimate: instead of an A/B test on copy, you can run five completely different behavioral variants.
Listen to this episode
Episode Description
Daniel Marcous, Co-Founder and CTO of April, describes how in a single weekend of vibe coding he and the VP R&D turned the accountants' review stage into 99% AI, and why he expects engineering velocity to grow by 1.3x to 1.5x over the coming year. The war with Iran, which sidelined much of the development team in Israel in the middle of tax season, became the catalyst for building a full multi-agent system in Cursor (architect, engineer, QA, PM and designer), which costs a power user about 1,000 to 2,000 dollars a month. Daniel explains which features go to engineers and which ones product managers, designers and even the CS and Finance teams build themselves all the way to production, why reading the agent's reasoning matters more than reading the code, and how he spots a builder mindset in interviews. He also shares the story of vibe coding with his five-and-a-half-year-old son, and his view that in a world where code is no longer a moat, a product strategy of many experiments in production becomes an advantage.
Chapters
- 00:00 · Meet Daniel and April (Calcalist's 50 most promising startups)
- 00:54 · DIY vs. human in the loop in the tax world
- 01:58 · The bottleneck that blew up on us last tax season
- 02:49 · The weekend we built an agent with vibe coding and made the review 99% AI
- 05:01 · The PM-to-engineer ratio has changed: 1:10 no longer holds
- 06:55 · Engineering velocity grows 1.3-1.5x: what it means for R&D
- 09:19 · How the Iran war became a catalyst for the next leap
- 11:31 · A brainstorm/hackathon every six months
- 12:00 · April's B2B2C business model and the pressure of tax season
- 14:00 · Building a multi-agent team: architect, engineer, QA, PM, designer
- 17:17 · The real cost: $1,000-$2,000 a month per power user
- 17:50 · How to work on 3 features in parallel
- 20:00 · What PMs can now do on their own: data analysis, sprint planning
- 22:11 · How CS and Finance at April build apps too
- 22:51 · Making sure PMs don't build gilded balconies
- 23:48 · From an A/B test of 2 variants to 5 complete variants
- 26:21 · Why the US tax market is a 30-year monopoly
- 28:08 · Vibe coding with Yonatan (Jon-Jon), age 5.5
- 32:24 · What a builder mindset is, and how to spot it
- 36:31 · The danger of gatekeepers and why you need to let go
- 38:38 · What a PM job interview at April looks like
- 50:00 · Tips for getting started with a team of agents
- 55:42 · Where the world of work is going, and the new philosophy of product strategy
Full transcript with speakers and timestamps
Selected quotes
“My next bottleneck is going to be product managers who can't produce precise requirements fast enough.”
“It's much, much more important to read the reasoning than the code it produces at the end.”
“The moat of code is already something that doesn't exist.”
Questions & answers
What does April do?
April builds an AI-based tax suite for American citizens, a kind of virtual accountant that handles tax planning, tax optimization and filing returns with the IRS and the states. Its business model is B2B2C, through financial institutions, and it also offers a track with a human accountant in the loop.
Why does Daniel Marcous plan to hire fewer engineers and more product managers?
Because AI tools are raising engineers' velocity, and he expects growth of 1.3x to 1.5x over the coming year. When engineers move faster, the bottleneck becomes product managers producing precise requirements fast enough, otherwise the development team ends up idle and working on technical debt.
How did April turn the accountants' review into 99% AI?
Over a weekend of vibe coding they built an agent that does a preliminary review: it takes the user's documents, runs the deterministic tax engine and the accountants' checklist, finds omissions and answers questions. After two days of iterating on the prompts, the accountants moved on to preparing returns.
How much does it cost to work with a team of agents?
According to Daniel, someone who uses the system intensively reaches 1,000 to 2,000 dollars a month, occasional use costs about 500 dollars, and light use about 100 dollars. He says the ROI on velocity easily justifies the cost.
How do you get started with a team of agents?
Beginners should build something simple in a domain they know, for example a game with their kids in Google AI Studio, to get past the courage barrier. The next step is Claude Code or Cursor: give the agent context and sources, agree on a plan with it, and test the multi-agent system on simple tasks before complex features.
People & resources mentioned
People
- Daniel Marcous
- Elon Musk
Companies & orgs
- April
- Calcalist
- Waze
- Intuit
- IRS
Tools
- Cursor
- Claude
- Claude Code
- Google AI Studio
- TurboTax
- BigQuery
- Slack
- Linear
- Figma
- Vercel
- Unity
Related Episodes
"I was afraid AI would take my job. It sharpened it instead" | Shira Moyal (imagen)
00:43:40 · May 10, 2026
Shira Moyal leads a product team at imagen, a startup that built a personal AI model which learns a professional photographer's editing style and edit...
אם אף אחד לא מתלונן, זה לא אומר שהכול טוב | בנצי ארבל (עסק מבפנים)
01:04:10 · November 10, 2026
אומרים שמי שרוצה לעשות כסף שלא ייכנס למסעדנות. ובכל זאת, בנצי ארבל, יחד עם השותף שלו עומר, בנה שלוש מסעדות, ביניהן כפרה מיו וקלבריה, ומפעל פסטה שמזין...
בוטסטראפ: איך בונים עסק שמתקיים מהכנסות?
00:33:49 · November 2, 2026
רוצים לבנות עסק טכנולוגי שההכנסות שלו מחזיקות אותו, במקום לרוץ מסיבוב גיוס אחד לשני? שיחה ליזמים בתחילת הדרך ולבעלי עסקי שירותים ששוקלים להפוך שירות ל...
From Zero to $2 Billion in Two Years, in a Saturated Market | Coffee with Eyal
00:08:21 · October 22, 2026
In the second episode of "Coffee with Eyal", Eyal David breaks down the story of David, a consumer company that jumped to a valuation of over 2 billio...