Apr 13, 2026 · 57m · latent-space
⚡️ The best engineers don't write the most code. They delete the most code. — Stay Sassy
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
Host Swix speaks with the anonymous creators of the tech publication 'Stay Sassy' to explore the operational realities of AI-assisted engineering, software development trade-offs, and why human empathy remains essential in technical leadership.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The hosts as informed peer 5.0 Guest teaching 3.6 Guest disagreement 1.9 The hosts pushing back 3.4
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 31:20 Dismissing leaf-node org chart automation
Stay Sassy EM dismisses typical executive automation plans as standard corporate nonsense, arguing that leaders mistakenly attempt to automate entry-level roles instead of their own routine bottlenecks.
Hardest push from the hosts ▶ 13:33 Direct challenge on AI token budgetsThe host bluntly interrupts the guest to state their $50k-$100k budget figures are far too low, offering concrete calculations from frontier AI researchers spending over $2.5 million annually.
Biggest teaching moment ▶ 17:25 Deconstructing the build vs buy illusionStay Sassy EM breaks down why engineers overestimate their ability to clone SaaS tools, citing unwritten PRD requirements, database migration risks, and long-term uptime maintenance.
The host holds their own ▶ 49:46 Countering code review necessity with the dark factory thesisThe host pushes back on the guests' insistence that teams must double down on manual pull reviews by introducing frontier research into automated dark factories where humans write and review no code.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Scaling Distribution Channels from Hacker News to Substack and X | 3 | 2 | 1 | 1 | Collaborative opening discussion where the host asks about growing an anonymous blog and the guests detail their progression across Hacker News, Substack, and X. | |
| Tech Satire, Viral Observations, and Skills Merch | 3 | 1 | 1 | 1 | Lighthearted discussion exploring the real-world inspiration behind the guests' viral satire tweets and the host's new merchandise concepts. | |
| Managing Individual AI Token Budgets and New Unit Economics | 7 | 3 | 2 | 6 | The host aggressively challenges the guests' token budgeting numbers, citing an unreleased interview with an OpenAI researcher spending 1 billion tokens daily. | |
| Re-evaluating Build Versus Buy Decisions in Software | 5 | 5 | 2 | 4 | The host tests internal tool rebuilding against his real conference SaaS costs, while the guests provide engineering management frameworks highlighting hidden operational complexity and key-man risk. | |
| Customizable UI Layers, Agent Workflows, and Cognitive Offloading | 6 | 3 | 2 | 4 | The host argues for personalizable UI layers over static backends, prompting a collaborative exploration of agentic interactions and consumer cognitive offloading. | |
| Automating Decision-Making and Executive Bottlenecks | 5 | 6 | 3 | 3 | Guests educate the host on why automation should target bottlenecked executives with standard decision trees rather than junior contributors at the leaves of the org chart. | |
| The Enduring Complexity of Human Management vs AI | 4 | 5 | 1 | 2 | Guests discuss the durable challenges of human management across startup stages and explain why AI managers fail to replace human trust. | |
| Code Quality, Review Fatigue, and Production Safety | 7 | 4 | 3 | 6 | The host contrasts the guests' defense of rigorous human code reviews against frontier concepts like the human-free 'software dark factory' and Latent Space's post on eliminating PRs. |
Statements from this episode (12)
Stay Sassy Hit Hacker News Front Page 5 to 10 Times Annually Early On
“And we had in his very early days, maybe like Five to 10 posts a year that we get like front page. And that was a huge sort of boost to our kind of early growth of the blog.”
Stay Sassy EM: Shift to API consumption pricing will redefine workplace management
“I think 20, 26 is going to be insane. It's already a little insane on that fund because we're seeing a lot of the big players move from kind of subsidy based request based pricing to straight up API pricing, consumption based pricing and I haven't seen a ton o…”
Stay Sassy EM: Companies can now build faster than they can distribute
“I think there are already businesses and we're seeing this in industry where they can build faster than they can distribute. They can build faster than they can get customers.”
Stay Sassy EM: Build internal software only if spreadsheet replaces it
“I think a very good for what it's worth, heuristic for some software is if you can replace it with a Google sheet or a spreadsheet. You probably can build it internally. And there's a lot of software that actually looks like a Google sheet or a spreadsheet wit…”
Stay Sassy EM: Dashboards must become agent-compatible this year
“The ability to do, like, any simple find-and-grab information from a dashboard, like, I think that just has to be agent-compatible, like, this year for companies, because that is one of those things where if you spend 10 minutes searching for a key that you ca…”
Stay Sassy PM: AI cognitive offloading will boost convenience services
“I would say, I think that this could be really good for any sort of service that makes, that allows people to be kind of lazy, because I think that AI has a very strong ability to just sort of reduce people's willingness to do work. It's like you could offload…”
Stay Sassy EM: 70% of executive work is standard and automatable
“I would start to think of the top of the tree and a lot of leaders and a lot of executives are bottlenecks for their entire organization. A lot of what they're doing, 70% of what they're doing should be things that Are pre-standard in industry.”
Stay Sassy PM: Leaders know what to do 99% of the time
“From what I've seen, 99% of the time, you know exactly what to do. The number of truly, truly difficult questions that require you to know some sort of advanced theory, or advanced philosophy, or run some huge amount of calculations that are very non straightf…”
Stay Sassy EM: AI startups face identical organizational problems as earlier generations
“I talked to people and they're like, Hey, I'm at like a series B hot AI startup and they have the exact same problems we had, you know, when we were at series, when I was at a series B and they had the exact same problems at series C. And so it's like it is It…”
Stay Sassy EM: A company will try an AI manager within two years
“And I really think that in the next two years, we will see the first company that tries an AI manager. They won't try an AI executive because I think that's like a whole different can of worms, but I think we'll see an AI manager and I think it'll be weird. An…”
Stay Sassy PM: Managing multiple AI agents exhausts engineers reviewing critical code
“One of the things that people quietly talk about is that that's very fatiguing. And I know some people who are operating that way. I've met these people and it is fatiguing. And what is fatigue really going to hit you on? Fatigue is really gonna hit you when n…”
Stay Sassy EM: Code review will bifurcate into full automation or rigorous discipline
“I think it's either going to go like PR is a totally automated review, and literally there's no human in the loop, and that the tools are good enough, or the entire industry is going to get to, we need to have poor review be like a serious thing that we're not…”