Apr 21, 2026 · 1h 0m · in-depth
Inside Artemis' "AI vs AI" war | Shachar Hirshberg & Dan Shiebler (Co-founders, Artemis)
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
Josh Koppelman interviews Artemis co-founders Shachar Hirshberg and Dan Shiebler about emerging from stealth with an AI-native threat detection platform, navigating the accelerating 'AI vs AI' cybersecurity landscape, and scaling a high-velocity startup.
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 →
Brett as informed peer 5.6 Guest teaching 4.9 Guest disagreement 1.4 Brett pushing back 2.5
speaking balance: gold is Brett, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 49:17 Rejecting the premise of slow legacy transition
Shachar directly challenges the notion that customers demand legacy parity bridges, citing that 70% completely replace old solutions immediately due to urgent threat vectors.
Hardest push from Brett ▶ 33:09 Questioning the scalability of texting customersJosh directly challenges Shachar on whether maintaining direct text threads with every CISO can realistically scale over a multi-year horizon.
Biggest teaching moment ▶ 50:51 Explaining fundamental control-layer architectureDan methodically breaks down for Josh why retrofitting legacy software with AI wrappers creates bottlenecked communication compared to native agentic control foundations.
Brett holds their own ▶ 50:31 Framing the AI-native vs AI-enabled divideJosh sharply articulates the core strategic dilemma facing enterprise buyers between incumbent AI additions and purpose-built startups, setting up Dan's architectural analysis.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Brett as informed peer | Guest teaching | Guest disagreement | Brett pushing back | Why |
|---|---|---|---|---|---|---|
| Defining Artemis and Building an AI-Native Product | 5 | 4 | 1 | 2 | Josh opens by setting up Artemis's trajectory and asking the founders to define the company and describe what production looks like. Shachar and Dan explain how their AI-native operational architecture delivers higher engagement compared to legacy SIEM tools. The dynamic is collaborative and conversational with minimal friction. | |
| Leveraging Deep Security Operations and Machine Learning Experience | 6 | 5 | 1 | 3 | Josh presses Shachar on the advantages and disadvantages of past domain expertise at AWS and Palo Alto, prompting Dan to explain the evolving threat landscape. Dan outlines how machine learning techniques from Abnormal apply to current behavioral detection problems. Josh actively probes with follow-ups like 'Like what type of things?'. | |
| Ideation Process, PMF Framework, and Co-Founder Partnership | 6 | 4 | 1 | 2 | Josh explores the ideation phase and co-founder chemistry, prompting Shachar to highlight the First Round PMF Method program. Shachar and Dan walk through their founder-market fit alignment and initial four-hour walk in the park. The exchange is supportive, highlighting shared investor-founder context. | |
| Timing the Market and Rethinking Startup Hiring | 5 | 6 | 2 | 3 | Josh drills into startup hiring practices and how Artemis tests for AI fluency. Dan explains their unconventional approach of weighting references heavily, running two-hour interview loops, and assessing end-to-end building capability over raw code syntax. The founders educate Josh on their specific screening heuristics. | |
| Engineering Architecture for AI Tooling and Automated Workflows | 5 | 6 | 2 | 3 | Dan and Shachar detail the technical architecture designed specifically to optimize AI agent performance and reduce hallucinations. Josh pushes on metrics outside engineering, prompting Dan and Shachar to explain how agentic product monitoring and automated telemetry replace manual analyst workflows. The founders walk through concrete operational examples. | |
| Identifying Core ICP and Early Product-Market Fit Signals | 6 | 5 | 1 | 2 | Josh asks how Artemis filtered through discovery calls to isolate their ideal customer profile. Shachar describes focusing strictly on upper mid-market and enterprise teams facing acute alert fatigue rather than casual experimenters. Dan explains how rapid data-source expansion serves as their core product-market fit signal. | |
| Founder-Led Sales and Strategic Stealth Emergence | 6 | 5 | 2 | 3 | Josh pushes on the scalability of high-touch texting relationships with CISOs as the sales organization scales. Dan and Shachar explain their strategy for codifying intuition into repeatable playbooks while maintaining founder ownership. Josh also questions how they managed to hire 30 people while remaining in stealth. | |
| In-Person Office Culture and Core Organizational Values | 5 | 4 | 1 | 2 | Josh questions the decision to establish an in-person culture in New York and asks how the co-founders manage strategic disagreements. Dan and Shachar articulate their five core values, noting how high velocity and high standards are unlocked through AI and strong mutual trust. The discussion is harmonious and philosophical. | |
| The AI vs AI Cybersecurity War and Architectural Moats | 7 | 6 | 2 | 3 | Josh raises a critical industry dilemma regarding how AI startups balance legacy feature parity with futuristic architectures. Shachar clarifies that 70% of clients outright replace legacy platforms due to escalating threats, while Dan gives a deep architectural breakdown of why AI-native control layers outperform AI-enabled legacy wrappers. | |
| Founder Mindset, Mentorship Influences, and Future Ambitions | 5 | 4 | 1 | 2 | Josh guides the closing segment on founder psychology, career timing, and pivotal mentorship figures. Shachar and Dan reflect on Demisto's founders and family influences while sharing lessons on managing flat organizations. Josh playfully points out their seed funding history before wrapping up on an appreciative note. |
Statements from this episode (19)
Hirshberg: Artemis reached 30 employees in seven months, adding 1-2 weekly
“We started seven months ago and we are now about 30 people in the team. Growing about one to two people every week and scaling across engineering product research and go to market.”
Hirshberg: Artemis users spend three to four hours daily in the platform
“What we see in practice is that the users of the product spend about three to four hours every day in the product because they are just able to get the outcomes they want using Artemis.”
Shiebler: Waiting six months longer would likely have made Artemis too late
“I think that in, in retrospect, if we'd waited An extra six months or so, as we were originally discussing, then probably we may have ended up being too late, given the shape of the market, given the speed in which things are developing.”
Hirshberg: Artemis co-founders spend over 60% of their time on hiring
“So we put, I would say over 60% of Dan's time and my time on hiring and developing the best talent in the world.”
Shiebler: Candidate reference strength perfectly correlates with actual job performance
“We found that to be extremely indicative. That really, really, we've, everybody, I think we can plot out on an almost, almost perfect correlation between the strength of references and the strength of which somebody actually delivers.”
Shiebler: Prior AI experience fails to predict workplace AI tool adoption
“I think that the, where somebody stands in terms of their prior experience at this point right now, given how long of a time that these tools have been around is, is not necessarily a great predictor of how much they'll lean into it and really adopt it.”
Shiebler: AI coding tools change software engineering physics by crossing domains
“AI coding tools enable engineers to be able to reach over the line and understand code that's outside of the range of things that they would normally be able to understand, normally be able to work with in a really unparalleled way that really changes the phys…”
Shiebler: Artemis evaluates all technical architecture by AI coding tool accuracy
“When we make architectural decisions, the first question that we ask is, is this an architectural decision that will increase or decrease the capability of AI tools to have the, to make the right answers?”
Hirshberg: Every Artemis engineer runs four to eight parallel cloud coding instances
“So today everyone is basically having like, let's say four to eight cloud code instances running in parallel, building the system, working on shipping four to five features simultaneously.”
Hirshberg: 50% of Artemis customers are in highly regulated industries
“And we tend to see about 50% of our customer base is highly regulated industries, such as financial services and financial institutions, where they have the need, the impact can be very high, but they don't necessarily have the Let's say desire to build in hou…”
Hirshberg: Artemis's first three customers asked to buy unprompted
“Reaching the level of trust and reaching the level of satisfaction where they actually came to us, the first three customers and told us that they want to buy before we ask them to buy the product.”
Hirshberg: Cybersecurity startups typically spend one to two years in stealth
“Typically, companies stay about a year to two years in stealth.”
Shiebler: Exiting stealth is for scaling sales and hiring, not initial customers
“You don't need to go out of stealth to delight your first customers. You need to go out of stealth to acquire additional customers and to be able to improve your ability to hire engineers.”
Hirshberg: Artemis extends engineering job offers within a day and a half
“So while other companies, their process will take three weeks, they already have an offer from us after a day and a half.”
Hirshberg: Artemis's in-person NYC office mandate has aided recruiting
“I think it helped. It's a self-selecting process, but we actually have a few people that were remote prior and they're missing human interaction. They're like, I want to be in the office. I want to be with others and brainstorm and whiteboard and just build to…”
Hirshberg: Within three years, AI agents will execute most global work
“When we think three years out in the future, we believe that most work will be done by agents. AI will direct the intent, sorry, humans will direct the intent, but AI will execute it. And that would go for probably most, let's say, work in the world, but speci…”
Hirshberg: 70% of Artemis customers fully replace legacy security solutions
“What we're seeing from customers, from even our customer base, about 70% of our customers replaced our legacy solutions with Artemis, because they say, I'm clear that at this point in time, I have to make the change right now in order to be prepared for what i…”