Nov 19, 2025 · 27m · latent-space

⚡️ 10x AI Engineers with $1m Salaries — Alex Lieberman & Arman Hezarkhani, Tenex

Arman Hezarkhani · 8m spoken Alex Lieberman · 8m spoken Shawn Wang · 6m spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

In this interview, Swyx speaks with Tenex co-founders Alex Lieberman and Arman Hezarkhani about how their AI-first engineering consultancy leverages output-based compensation to reward 10x developers earning up to $1 million. They detail their structured technical stack, rigorous hiring philosophy, and entropy management in autonomous coding loops ahead of the AI Engineer Conference.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 25.4% of the talking time here. How this is scored →

The hosts as informed peer 4.3 Guest teaching 3.0 Guest disagreement 2.7 The hosts pushing back 3.3
05100:0010:0020:000:03–5:42 · The hosts as informed peer 2/10 Introductions and Name Pronunciation Banter The conversation opens with lighthearted rapport over name pronunciation before Alex provides the foundational backstory of Tenex and its output-based compensation model. Swyx acts primarily as a receptive host guiding the intro.5:42–8:33 · The hosts as informed peer 4/10 Story Point Compensation and Aligning Client Incentives Swyx immediately expresses skepticism about measuring developer output via story points. Arman and Alex counter by explaining their dual incentives model, balancing long-term developer incentives against account-level technical strategists.8:33–12:20 · The hosts as informed peer 2/10 High Compensation Milestones and Rapid Prototyping Case Studies Swyx prompts the guests to brag about high compensation and successful case studies. Arman highlights million-dollar engineer earnings and rapid edge-AI deployment, while Alex describes rapid prototyping in sales motions.12:21–16:02 · The hosts as informed peer 5/10 Tenex Engineering Stack and Talent Scaling Constraints Swyx challenges Tenex's dynamic coding agent selection as being overly anecdotal rather than rigorous and eval-driven. Arman pushes back with a samurai sword analogy, arguing that at the frontier, agent selection comes down to qualitative feel.16:02–21:53 · The hosts as informed peer 6/10 Vetting AI Engineers and the Autonomous Entropy Problem The guests outline their rigorous take-home tests and bring engineer Dan on to explain how autonomous agent loops fail due to compounding entropy. Swyx engages actively, linking Dan's concept to known failure modes in context engineering.21:54–25:58 · The hosts as informed peer 7/10 Navigating AI Concepts and Debating Model Context Protocol Arman dismisses Model Context Protocol (MCP) as a re-hyped acronym for basic APIs, prompting Swyx to defend the protocol's broader technical spec and explain the sociological necessity of new terminology.0:03–5:42 · Guest teaching 2/10 Introductions and Name Pronunciation Banter The conversation opens with lighthearted rapport over name pronunciation before Alex provides the foundational backstory of Tenex and its output-based compensation model. Swyx acts primarily as a receptive host guiding the intro.5:42–8:33 · Guest teaching 3/10 Story Point Compensation and Aligning Client Incentives Swyx immediately expresses skepticism about measuring developer output via story points. Arman and Alex counter by explaining their dual incentives model, balancing long-term developer incentives against account-level technical strategists.8:33–12:20 · Guest teaching 3/10 High Compensation Milestones and Rapid Prototyping Case Studies Swyx prompts the guests to brag about high compensation and successful case studies. Arman highlights million-dollar engineer earnings and rapid edge-AI deployment, while Alex describes rapid prototyping in sales motions.12:21–16:02 · Guest teaching 3/10 Tenex Engineering Stack and Talent Scaling Constraints Swyx challenges Tenex's dynamic coding agent selection as being overly anecdotal rather than rigorous and eval-driven. Arman pushes back with a samurai sword analogy, arguing that at the frontier, agent selection comes down to qualitative feel.16:02–21:53 · Guest teaching 5/10 Vetting AI Engineers and the Autonomous Entropy Problem The guests outline their rigorous take-home tests and bring engineer Dan on to explain how autonomous agent loops fail due to compounding entropy. Swyx engages actively, linking Dan's concept to known failure modes in context engineering.21:54–25:58 · Guest teaching 2/10 Navigating AI Concepts and Debating Model Context Protocol Arman dismisses Model Context Protocol (MCP) as a re-hyped acronym for basic APIs, prompting Swyx to defend the protocol's broader technical spec and explain the sociological necessity of new terminology.0:03–5:42 · Guest disagreement 1/10 Introductions and Name Pronunciation Banter The conversation opens with lighthearted rapport over name pronunciation before Alex provides the foundational backstory of Tenex and its output-based compensation model. Swyx acts primarily as a receptive host guiding the intro.5:42–8:33 · Guest disagreement 2/10 Story Point Compensation and Aligning Client Incentives Swyx immediately expresses skepticism about measuring developer output via story points. Arman and Alex counter by explaining their dual incentives model, balancing long-term developer incentives against account-level technical strategists.8:33–12:20 · Guest disagreement 1/10 High Compensation Milestones and Rapid Prototyping Case Studies Swyx prompts the guests to brag about high compensation and successful case studies. Arman highlights million-dollar engineer earnings and rapid edge-AI deployment, while Alex describes rapid prototyping in sales motions.12:21–16:02 · Guest disagreement 4/10 Tenex Engineering Stack and Talent Scaling Constraints Swyx challenges Tenex's dynamic coding agent selection as being overly anecdotal rather than rigorous and eval-driven. Arman pushes back with a samurai sword analogy, arguing that at the frontier, agent selection comes down to qualitative feel.16:02–21:53 · Guest disagreement 3/10 Vetting AI Engineers and the Autonomous Entropy Problem The guests outline their rigorous take-home tests and bring engineer Dan on to explain how autonomous agent loops fail due to compounding entropy. Swyx engages actively, linking Dan's concept to known failure modes in context engineering.21:54–25:58 · Guest disagreement 5/10 Navigating AI Concepts and Debating Model Context Protocol Arman dismisses Model Context Protocol (MCP) as a re-hyped acronym for basic APIs, prompting Swyx to defend the protocol's broader technical spec and explain the sociological necessity of new terminology.0:03–5:42 · The hosts pushing back 1/10 Introductions and Name Pronunciation Banter The conversation opens with lighthearted rapport over name pronunciation before Alex provides the foundational backstory of Tenex and its output-based compensation model. Swyx acts primarily as a receptive host guiding the intro.5:42–8:33 · The hosts pushing back 5/10 Story Point Compensation and Aligning Client Incentives Swyx immediately expresses skepticism about measuring developer output via story points. Arman and Alex counter by explaining their dual incentives model, balancing long-term developer incentives against account-level technical strategists.8:33–12:20 · The hosts pushing back 1/10 High Compensation Milestones and Rapid Prototyping Case Studies Swyx prompts the guests to brag about high compensation and successful case studies. Arman highlights million-dollar engineer earnings and rapid edge-AI deployment, while Alex describes rapid prototyping in sales motions.12:21–16:02 · The hosts pushing back 5/10 Tenex Engineering Stack and Talent Scaling Constraints Swyx challenges Tenex's dynamic coding agent selection as being overly anecdotal rather than rigorous and eval-driven. Arman pushes back with a samurai sword analogy, arguing that at the frontier, agent selection comes down to qualitative feel.16:02–21:53 · The hosts pushing back 3/10 Vetting AI Engineers and the Autonomous Entropy Problem The guests outline their rigorous take-home tests and bring engineer Dan on to explain how autonomous agent loops fail due to compounding entropy. Swyx engages actively, linking Dan's concept to known failure modes in context engineering.21:54–25:58 · The hosts pushing back 5/10 Navigating AI Concepts and Debating Model Context Protocol Arman dismisses Model Context Protocol (MCP) as a re-hyped acronym for basic APIs, prompting Swyx to defend the protocol's broader technical spec and explain the sociological necessity of new terminology.

speaking balance: gold is the hosts, purple is the guest (3 minute bins)

0:00 · the hosts 22.3% · guest 77.7%0:00 · the hosts 22.3% · guest 77.7%3:00 · the hosts 17.2% · guest 82.8%3:00 · the hosts 17.2% · guest 82.8%6:00 · the hosts 19.3% · guest 80.7%6:00 · the hosts 19.3% · guest 80.7%9:00 · the hosts 0% · guest 100%9:00 · the hosts 0% · guest 100%12:00 · the hosts 18.4% · guest 81.6%12:00 · the hosts 18.4% · guest 81.6%15:00 · the hosts 15.3% · guest 84.7%15:00 · the hosts 15.3% · guest 84.7%18:00 · the hosts 21.2% · guest 78.8%18:00 · the hosts 21.2% · guest 78.8%21:00 · the hosts 46.7% · guest 53.3%21:00 · the hosts 46.7% · guest 53.3%24:00 · the hosts 69.8% · guest 30.2%24:00 · the hosts 69.8% · guest 30.2%27:00 · the hosts 0% · guest 0%27:00 · the hosts 0% · guest 0%
Sharpest disagreement ▶ 23:46 Arman brands MCP as an overhyped acronym for APIs

Arman directly mocks the industry hype cycle around MCP, calling it a glorified three-letter acronym used to raise venture capital and generate viral alarmist tweets.

Hardest push from the hosts ▶ 5:42 Swyx challenges output measurement and Goodhart's law

Swyx openly tells the guests he is default skeptical of measuring engineering output by story points because anything measured gets gamed.

Biggest teaching moment ▶ 20:07 Dan breaks down agent failure through compounding entropy

Dan steps in and reframes the bottleneck of autonomous engineering from pure context size to mathematical error compounding across agent loops.

The host holds their own ▶ 25:10 Swyx defends the MCP specification against reductionism

Swyx demonstrates his technical depth by pointing out that critics ignore parts of the MCP specification beyond basic API wrappers, defending its architectural merits.

the scores for every segment, with the reasoning behind each
ChapterTopicThe hosts as informed peerGuest teachingGuest disagreementThe hosts pushing backWhy
Introductions and Name Pronunciation Banter 2211 The conversation opens with lighthearted rapport over name pronunciation before Alex provides the foundational backstory of Tenex and its output-based compensation model. Swyx acts primarily as a receptive host guiding the intro.
Story Point Compensation and Aligning Client Incentives 4325 Swyx immediately expresses skepticism about measuring developer output via story points. Arman and Alex counter by explaining their dual incentives model, balancing long-term developer incentives against account-level technical strategists.
High Compensation Milestones and Rapid Prototyping Case Studies 2311 Swyx prompts the guests to brag about high compensation and successful case studies. Arman highlights million-dollar engineer earnings and rapid edge-AI deployment, while Alex describes rapid prototyping in sales motions.
Tenex Engineering Stack and Talent Scaling Constraints 5345 Swyx challenges Tenex's dynamic coding agent selection as being overly anecdotal rather than rigorous and eval-driven. Arman pushes back with a samurai sword analogy, arguing that at the frontier, agent selection comes down to qualitative feel.
Vetting AI Engineers and the Autonomous Entropy Problem 6533 The guests outline their rigorous take-home tests and bring engineer Dan on to explain how autonomous agent loops fail due to compounding entropy. Swyx engages actively, linking Dan's concept to known failure modes in context engineering.
Navigating AI Concepts and Debating Model Context Protocol 7255 Arman dismisses Model Context Protocol (MCP) as a re-hyped acronym for basic APIs, prompting Swyx to defend the protocol's broader technical spec and explain the sociological necessity of new terminology.

Statements from this episode (7)

Insight
Lieberman: Hourly billing perversely penalizes workers who use AI faster
“And also you're perversely incentivized because you leveraging AI in your work as you operating faster, but your incentive, just like a lawyer or just like any hour pay based knowledge worker is to rack up as many hours as possible.”
Alex Lieberman Nov 19, 2025 ▶ 4:08
Insight
Hezarkhani: Traditional startup employees lack incentives to maximize AI productivity
“Even if you have some equity, even if you deeply care about the mission, you're not deeply incentivized day in and day out to try new AI tools and push yourself to work better and faster and smarter.”
Arman Hezarkhani Nov 19, 2025 ▶ 5:01
Prediction Not checkable as stated
Hezarkhani: Tenex will have engineers making over $1M cash next year
“We will probably have more than one engineer make million dollars cash next year based on this model. And that is just with story point compensation. It's very likely that we will have more than a handful of folks make more than a million dollars next year.”
Arman Hezarkhani Nov 19, 2025 ▶ 9:03
Insight
Hezarkhani: Structured TypeScript architectures allow AI agents to operate autonomously for longer
“We feel pretty strongly in like high structure allows for agents to work autonomously for longer. And so our default stack is TypeScript front end TypeScript back end with a shared file where, or a shared folder where all of our shared types and schemas and th…”
Arman Hezarkhani Nov 19, 2025 ▶ 12:38
Disclosure
Lieberman: Tenex's business growth is 100% human talent constrained, not agent constrained
“Today it's human bound, a hundred percent.”
Alex Lieberman Nov 19, 2025 ▶ 15:12
Opinion
Hezarkhani: MCP is essentially just a three-letter word for API
“I just think that MCP is a three-letter word for API”
Arman Hezarkhani Nov 19, 2025 ▶ 23:47
Opinion
Swyx: MCP specification includes features beyond simple API wrappers
“I will defend MCP in the sense that, like, there actually are other parts of the spec that are not just API wrappers, but people just comparatively don't use them as much, but I think it's a little unfair to MCP, the whole protocol”
Shawn Wang Nov 19, 2025 ▶ 25:11
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 200 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.