May 7, 2026 · 59m · big-technology
AI Agents: Mirage Or Real Revolution? — With Dmitry Shevelenko
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 episode of the Big Technology Podcast, host Alex Kantrowitz interviews Dmitry Shevelenko, Chief Business Officer at Perplexity, to examine the real-world viability, economic models, and competitive strategies behind agentic AI super apps. Shevelenko explains how model-agnostic orchestration, consumption-based pricing, and dedicated workflow tools are transforming generative AI from fleeting novelty into high-leverage digital labor.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 32.5% of the talking time here. How this is scored →
Alex as informed peer 4.9 Guest teaching 4.5 Guest disagreement 2.3 Alex pushing back 3.4
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 4:59 Dismissing Traffic Flatlining with 500M ARR Metric
Shevelenko dismisses Kantrowitz's cited web analytics by asserting revenue growth has crossed 500M ARR, arguing traffic metrics are irrelevant for a non-advertising utility business.
Hardest push from Alex ▶ 7:12 Insisting Active User Growth Precedes MonetizationKantrowitz refuses Shevelenko's pivot to revenue, challenging that tech companies only focus on revenue metrics after active user expansion hits a ceiling.
Biggest teaching moment ▶ 51:37 Educating Host on Chinese Chip-Level Architecture Lock-InShevelenko educates Kantrowitz on why Jensen Huang is protective of the Western hardware stack, explaining that software architectures could be designed to run exclusively on Huawei silicon rather than Nvidia.
Alex holds their own ▶ 3:24 Drilling into Perplexity Traffic and AI App PlateauKantrowitz cites detailed market share and daily active visit statistics from Apptopia and SimilarWeb to build a rigorous case that consumer search AI has plateaued.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Apple Acquisition Speculation and the Enduring Value of Hardware | 5 | 3 | 2 | 3 | Kantrowitz opens by recalling his public prediction that Apple would acquire Perplexity and probes about Apple's incoming leadership. Shevelenko politely jokes about the prediction failing and highlights their growing Mac Mini infrastructure collaboration and independence. | |
| Consumer AI Usage Plateau vs. Perplexity's Revenue Growth | 7 | 5 | 4 | 6 | Kantrowitz presses Shevelenko with third-party web traffic and market share data from Apptopia and SimilarWeb showing consumer AI plateauing. Shevelenko counters by rejecting top-line traffic metrics in favor of doubling ARR to 500 million dollars, prompting Kantrowitz to push back that MAU growth remains fundamental. | |
| Deconstructing the Consumer AI Slowdown and Capabilities Overhang | 4 | 5 | 2 | 2 | Kantrowitz explores the broader stagnation of consumer AI usage across the industry. Shevelenko explains the capabilities overhang, noting consumers still use generative tools for basic web information retrieval rather than high-leverage workflows. | |
| Novelty Cycles, Multimodal Spikes, and Multi-Agent Leverage | 6 | 3 | 1 | 2 | Kantrowitz shares his thesis on multimodal novelty spikes such as voice and image generation trailing off into plateau periods. Shevelenko agrees and uses Kantrowitz's independent media production setup as an example of real economic leverage from multi-agent orchestration. | |
| Perplexity Computer: Shifting from Software Budgets to Digital Payroll | 5 | 4 | 3 | 5 | Kantrowitz questions whether computer agents might just be another ephemeral novelty cycle like early chat tools. Shevelenko defends the product model, arguing users treat agent compute credits as payroll budget for digital labor rather than discretionary software spending. | |
| Automating High-Value Workflows and Fact-Checking with Final Pass | 4 | 4 | 2 | 3 | Kantrowitz questions the reliability of trusting autonomous agents with high-stakes tasks like accounting and tax filing. Shevelenko reframes the capability around automated auditing and fact-checking human errors via their Final Pass workflow. | |
| Granular System Permissions, Security Safeguards, and Mac Mini Infrastructure | 5 | 6 | 3 | 5 | Kantrowitz lists the intrusive permissions required to connect Perplexity Computer to email and calendar, questioning if running on Mac Minis is an isolation sandbox. Shevelenko clarifies that Mac Minis are meant for continuous 24/7 background execution and deeper local app access rather than local air-gapping. | |
| Perplexity's Moat: Agnostic Model Orchestration, Search Grounding, and Usability | 4 | 5 | 2 | 2 | Kantrowitz asks how Perplexity can defend its agent product against OpenAI Codex and Anthropic Claude Code. Shevelenko details their structural moats in multi-model aggregation, grounded search index flywheels, and workflow-oriented interface design. | |
| Model Provider Platform Dynamics, Competitive Parity, and Frontier Security | 6 | 4 | 3 | 5 | Kantrowitz presses on the vulnerability of relying on upstream foundation model providers who are building competing downstream agent apps. Shevelenko counters that frontier model parity and fierce API platform competition prevent labs from restricting access. | |
| Chinese Open-Source Models, Post-Training Safeguards, and Hardware Sovereignty | 5 | 7 | 3 | 4 | Kantrowitz queries why national origin matters if open-source model weights like Kimi K2 can be downloaded and post-trained locally. Shevelenko breaks down the strategic risk of open models being optimized strictly for Huawei chips over Nvidia hardware, earning immediate recognition from Kantrowitz. | |
| Subsidized Pricing Distortions and Perplexity's Costco Consumption Model | 5 | 5 | 2 | 3 | Kantrowitz brings up reporting that AI market demand is artificially inflated by subsidized flat-rate subscription tiers. Shevelenko explains Perplexity's non-subsidized credit consumption model, comparing it to Costco's membership access plus metered utility. | |
| Maintaining Lean Organizational Agility and Show Conclusion | 3 | 3 | 1 | 1 | Kantrowitz asks how Perplexity navigates rapid product shifts in an unstable technological landscape. Shevelenko emphasizes small headcount, organizational flexibility, and rapid decision iteration to close the interview. |
Statements from this episode (21)
Shevelenko: Perplexity has a blossoming partnership with Apple using Mac minis
“We have a great blossoming partnership with Apple. They actually are really excited about what we're doing with personal computer and how it uses Mac minis.”
Shevelenko: Hardware will matter more as software faces commoditization pressure
“Apple's always been an incredible hardware company, and I think, ah, this is, you know, an era where hardware will matter even more because software is gonna face waves of commoditization pressure.”
Shevelenko: Computer-using AI agents were impossible before late 2025
“You couldn't have built something like computer before November, December of last year, because model capabilities advance where you can have longer time horizons for running tasks, right? Where you're not just answering a question, but you're actually doing W…”
Shevelenko: Ads alongside AI search answers undermine perceived accuracy
“When a core value prop of perplexity is accuracy it's really hard to reinforce that to users when you also have ads running alongside the answer.”
Kantrowitz: Anthropic and OpenAI will both have trillion-dollar IPOs
“We're looking at, they're both, Anthropik and OpenAI are both gonna have trillion dollar IPOs, and we'll have many large companies, I think, that will follow them in the generative AI world.”
Kantrowitz: ChatGPT reached 200 million users on text alone
“ChatGPT got to two hundred million users because of text.”
Shevelenko: Perplexity Computer users consume more credits each week, driving revenue
“The longer people have had access to computer. I mean, this stuff is still brand new, but they're using it consuming more computer credits every week. Than the previous week, right? So we're actually just in, in the extreme upward part of the ramp. That's a bi…”
Shevelenko: Perplexity launching 36 workflows on top of Perplexity Computer
“So we actually are launching this week, 36 different workflows that go on top of computer.”
Perplexity content team ships production code without engineers via Perplexity Computer
“We have people on our content team that submit pull requests, basically ship You know, code that goes live into production without engineers being in the loop. And that's all being run through Perplexity Computer.”
Shevelenko predicts powerful reasoning models will run locally on Mac Minis
“Well, I certainly think that as models get more powerful, you will certainly be, and as, you know, local CPUs get more powerful as well, you're going to be able to distill powerful reasoning models to a size where they can run on a Mac mini.”
Shevelenko sees no indication the AI data center buildout is a bubble
“From the perplexity point of view, like we don't have strong opinions on the data center build out, but there's nothing I see that indicates that that is, you know, a bubble or anything like that.”
Shevelenko: Codex and proprietary AI assistants will never run rival models
“So the one thing that Codex is never going to be able to support is running Gemini models. You know, they will always be in the GPT family. Same thing for, you know, Claude, like they're not gonna, you know, have GPT models. Gemini is not gonna have Grok model…”
Shevelenko: Non-AI companies gain alpha from adoption depth, not custom tools
“The alpha for a company that is not an AI company is, is not in them building their own internal tools with AI necessarily. It is in the depth of their adoption.”
Shevelenko: AI labs aggressively petition Perplexity to use their models
“They aggressively petition us to use their models. They want, they give us early access. They want us to run evals. And so we have, you know, the exact opposite dynamic right now where they're, you know, more than happy to take revenue from us.”
Shevelenko: The gap between the top two AI models has never exceeded 15%
“Since this industry has kicked off, there's never been a moment where the delta between the, you know, the best model and the second best model was, like, more than maybe, like, a 10, 15% gap.”
Shevelenko: Models may be better at exploiting cyber vulnerabilities than fixing them
“I think what is a real concern is that models will be better at exploiting cyber vulnerabilities than they are at fixing them, right?”
Shevelenko: Perplexity hosts and post-trains Chinese open-source models in US data centers
“We never Integrate into perplexity any product that, or API that is hosted in China. We have ourselves post-trained. We have post-trained open source models. That, that come that are developed by Chinese labs. We run those in US data centers. We post-train the…”
Shevelenko: Chinese AI models are pushing the frontier but not at it
“They are, you know, they're pushing the frontier. They're not at the frontier, but they're pushing it.”
Shevelenko: Perplexity has never subsidized paying Pro or Max users
“So we at Perplexity, we've never subsidized paying users. So if you're on a, you know, pro or max plan you know, we're, thank you, you're contributing to our success.”
Shevelenko: AI pricing will resemble Costco with memberships and compute credits
“AI is going to become a lot like Costco where you pay for the membership, right? And that gets you in the store. And that's actually the part of Costco's business that is, you know, the highest margin. And then you have, you know, everything you're buying in t…”