Dec 23, 2025 · 29m · big-technology
Resolve AI CEO Spiros Xanthos: AI for Prod, Multi-agent Architectures, Engineering's Future
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
Resolve AI CEO Spiros Xanthos explains why rapid AI code generation creates severe production bottlenecks and how multi-agent architectures and autonomous operational agents solve them. He explores the tiered evolution of production autonomy, model orchestration hierarchies, and how AI elevates software engineers to focus on architectural strategy.
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 30.3% of the talking time here. How this is scored →
Alex as informed peer 4.7 Guest teaching 4.4 Guest disagreement 1.0 Alex pushing back 1.4
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 25:32 Dismissing Developer Skill Atrophy
Xanthos counters the host's premise about engineering skill loss, arguing that software has continuously moved to higher abstraction layers over fifty years without issue.
Hardest push from Alex ▶ 16:50 Challenging Multi-Agent FragilityKantrowitz directly challenges agentic architectures by noting that if one agent fails in a chained workflow, the entire system collapses.
Biggest teaching moment ▶ 17:18 Explaining Supervisor Validation LoopsXanthos educates the host on how production multi-agent systems prevent failures by stacking peer review agents and supervisor feedback loops.
Alex holds their own ▶ 5:45 Dissecting the ROI Paradox of AI CodeKantrowitz articulates the operational trade-off where rapid code generation diminishes enterprise ROI by creating massive downstream monitoring workloads.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Data Flywheels and Feedback Loops in Software | 5 | 4 | 1 | 1 | Kantrowitz highlights the difference between deterministic coding reward loops and open-ended disciplines. Xanthos confirms the premise and elaborates on how developer IDE acceptance data builds a continuous flywheel. | |
| The Production Challenge of AI-Generated Code | 5 | 4 | 1 | 1 | The host identifies the production bottleneck where increased AI code generation diminishes net ROI by multiplying maintenance burdens. The guest agrees, explaining that unmonitored code becomes an operational liability. | |
| Levels of Autonomy in Production Engineering | 4 | 5 | 1 | 2 | Kantrowitz presses on the actual level of autonomy granted to AI tools during production outages. Xanthos uses an autonomous vehicle analogy to describe the phased progression toward autonomous remediation. | |
| Multi-Agent Systems and Domain Knowledge | 6 | 5 | 1 | 2 | The host questions model commoditization and suggests multi-model orchestration is where frontier progress is happening. Xanthos expands on long-horizon reasoning and capturing tribal institutional knowledge. | |
| Multi-Agent Orchestration and Model Hierarchies | 5 | 5 | 1 | 2 | Kantrowitz questions multi-agent failure cascades and model routing tiers. Xanthos explains supervisor architectures, spot checking, and pairing proprietary frontier orchestrators with specialized task agents. | |
| Engineering Culture and the Evolution of Developer Skills | 5 | 4 | 2 | 2 | The host addresses concerns regarding software engineering skill atrophy when AI writes and audits code. Xanthos reframes AI as an evolutionary layer of abstraction rather than a threat to developer capabilities. | |
| Real-World Incident Remediation in Production | 3 | 4 | 0 | 0 | Kantrowitz asks for an end-to-end incident remediation case study. Xanthos details how the platform isolates end-user bugs, walks microservice infrastructure, and recommends rolling back specific commits. |
Statements from this episode (13)
Xanthos: AI model progress has not stalled and remains exponential
“I do believe we're still in an exponential improvement curve with AI. I think the models keep improving quite a bit because there was maybe a concern, maybe a few months ago, more like, you know, end of last year, I would say a year ago, whether let's say the …”
Xanthos: Agentic AI will expand across business processes in 2026
“And I think what we will start to seeing now in 20, 26, that paradigm is going to show up in other parts of software and in other industries where they see the customer service and a lot of other business process automation.”
Xanthos: Coding AI's human acceptance feedback flywheel applies to other domains
“I do think that paradigm is applicable to other domains. So you have to probably start with something that is useful, that let's say humans are on the driver's seat, engineers in our case, and As long as it's a better way of working, it's faster than not witho…”
Resolve AI's primary use case is autonomous on-call incident remediation
“The primary way engineers use Resolve AI Is to put us on call instead of them waking up in the middle of the night when something goes wrong with software and resolve does the triaging and troubleshooting and suggests remediation.”
Xanthos: More code without downstream support becomes a corporate liability
“A lot more code without addressing the subsequent steps is almost a liability. It's not an asset for a company, right?”
Xanthos: Studies show production incidents rising alongside AI coding tool adoption
“So there are studies out there that showed, let's say, that the incidents where something went wrong, their new, new change has increased quite a bit as AI is being used.”
Xanthos: AI will primarily drive software development within one year
“I do believe that, you know, we are probably a year away from AI becoming the driver of software, the same way, let's say, agents are the primary producer of code today. I think we're going to move to the same place in a year from now, where humans are going t…”
Xanthos: AI will make most engineering decisions in two to three years
“Still overseeing that AI, making most of the final decisions, but I think probably in two to three years, we're going to be at the place where AI is going to be making most of these decisions, and humans will be delegating, let's say, however, maybe decision f…”
Xanthos: AI products must integrate institutional tribal knowledge to succeed
“I do believe that to be very successful, and I've seen many examples of that, you need to incorporate a lot of the knowledge, specifics, the tribal knowledge or institutional knowledge into, let's say, the product or the model to be able to be highly effective…”
Xanthos: Multi-agent AI needs giant models for orchestration, smaller models below
“At the top level, you usually have the most capable model, usually the most expensive also, and the biggest model, and then the underlying tasks can be performed by maybe either fast closed source models, or even open source models that you posturing for the t…”
Xanthos: Domain-specialized AI models will ultimately outperform general horizontal models
“I do think we're going to end up in a situation where the most capable model in a domain is going to be a specialized model.”