In this episode, we explore how a startup called Edgeseek uses eBPF to run SQL queries directly in the Linux kernel, bypassing traditional database layers for ultra-low-latency packet filtering and observability. We break down the technical architecture — how they compile SQL to BPF bytecode, attach it to kernel hooks, and stream results back to user space at line rate — and discuss the trade-offs: limited memory, no joins, no disk persistence. If you're building high-throughput data pipelines or wondering where the boundary between databases and operating systems is blurring, this episode is for you.

#eBPF #SQL #Kernel #Edgeseek #LowLatency #Observability #PacketFiltering #BPF #Linux #DataPipeline #HighThroughput #Startup #Engineering #Technology #BusinessAndTechnology #FexingoBusiness #BusinessPodcast #TechnicalCoFounder

Keep every episode free: buymeacoffee.com/fexingo

In this episode, Lucas and Luna explore how a fast-growing fintech startup, PayFlow, uses Falco — an open-source runtime security tool originally built by Sysdig — to detect and block container-level threats in real time. They walk through a specific incident where Falco's default rules caught a cryptominer deployment within seconds, saving the company from a potential six-figure cloud bill. The conversation covers how PayFlow customized Falco's rule engine to reduce false positives, integrated it with their existing Prometheus alerting stack, and why runtime security is becoming a must-have for any team running Kubernetes in production. Lucas explains the difference between image scanning and runtime detection, and Luna pushes back on whether Falco is practical for smaller teams. The episode is anchored in a concrete deployment scenario that any engineering leader running containers can relate to.

#Falco #RuntimeSecurity #ContainerSecurity #Kubernetes #CloudNative #DevSecOps #Fintech #PayFlow #Sysdig #Cryptomining #KubernetesSecurity #ContainerThreats #Prometheus #OpenSource #BusinessAndTechnology #TheTechnicalCoFounderPodcast #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

On today's episode, Lucas and Luna explore how a startup called PromptLab uses feature flags to A-B test different AI prompt templates in production — without redeploying code. They break down the technical architecture: a custom prompt router that checks a LaunchDarkly-style flag, serves variant A or B, logs token usage and output quality, then feeds results into a feedback loop for prompt engineers. Lucas walks through the concrete tradeoffs — caching prompt templates in a Redis sidecar, using semantic similarity to detect regressions, and the decision to build their own prompt management layer versus buying an LLM observability platform. Luna asks whether A-B testing prompts is fundamentally different from A-B testing UI, and why most companies don't do it well. They tie it back to the broader shift: as LLM features become core product surfaces, engineering teams need the same rigor for prompt versioning as they have for code versioning. A sharp, practical episode for anyone shipping AI features.

#FeatureFlags #A-BTesting #AIPrompts #LLM #PromptEngineering #LaunchDarkly #PromptVersioning #SoftwareEngineering #StartupTech #ProductionTesting #ModelObservability #RigorousEngineering #TechPodcast #FexingoBusiness #BusinessPodcast #AIEngineering #PromptLab #TechnicalCoFounder

Keep every episode free: buymeacoffee.com/fexingo

Episode 92 of The Technical Co-Founder Podcast dives into the concept of the 'paved road' — a pre-configured, opinionated platform that lets developers ship features without reinventing infrastructure. Lucas and Luna examine how a mid-stage B2B SaaS startup called Flowtile reduced its onboarding-to-first-deploy time from two weeks to under four hours by building an internal developer platform around a standardised container runtime, CI/CD pipeline, and observability stack. The episode walks through the concrete decisions: why they chose a single container orchestration provider, how they baked in security and cost guardrails, and the surprising cultural pushback from senior engineers who felt the paved road limited their autonomy. Flowtile's CTO, Jenna Kwan, shares that the paved road cut their P50 deploy time from 18 minutes to 8 minutes and eliminated three categories of production incidents. The hosts also touch on the trade-off between flexibility and consistency, and how Flowtile measures developer experience through a quarterly 'velocity score' tied to a data-backed metrics framework.

#PavedRoad #DeveloperVelocity #InternalDeveloperPlatform #IDP #DevEx #Flowtile #JennaKwan #ContainerRuntime #CI/CD #PlatformEngineering #StartupEngineering #TechLeadership #DeveloperProductivity #CTO #GoldenPath #FexingoBusiness #BusinessPodcast #BusinessAndTechnology

Keep every episode free: buymeacoffee.com/fexingo

Lucas and Luna dig into why a growing number of engineering teams are rewriting performance-critical Python services in Rust. They break down the specific case of a real-time analytics startup that cut processing time from 12 seconds to under 200 milliseconds by porting its core data pipeline to Rust. The hosts discuss the trade-offs: Rust's steep learning curve and slower initial development vs. Python's ease and ecosystem. They talk about when it makes sense to reach for Rust — not everywhere, but in hot-path services where latency and memory matter. Luna shares a story from a fintech CTO who rewrote just one Python service and saved 40% on cloud costs. Lucas explains why Rust's ownership model eliminates memory bugs at compile time. They also touch on tooling like PyO3 for incremental integration, and the debate over microservices vs. a hybrid approach. If you're a technical founder or CTO weighing the cost-benefit of a language migration, this episode is for you.

#Rust #Python #Startup #Engineering #Performance #Rewriting #CTO #TechnicalCoFounder #FexingoBusiness #BusinessPodcast #TechPodcast #Backend #SystemsProgramming #PyO3 #OwnershipModel #CloudCosts #Latency #MemorySafety

Keep every episode free: buymeacoffee.com/fexingo

Lucas and Luna dive into the story of a logistics startup that replaced nightly batch ETL jobs with serverless SQL queries against their data lake. The founders were struggling with a 24-hour delay in shipment tracking data. By switching to Athena and Presto on S3, they cut query latency from hours to seconds and reduced their data infrastructure bill by 60%. Lucas explains the technical architecture: partitioned Parquet files, columnar compression, and how they used AWS Glue for schema inference. Luna asks the hard questions about when this approach breaks down—like concurrency limits and hot partitions. They also discuss the trade-offs between serverless SQL and traditional data warehouses like Redshift or Snowflake. If you're building analytics on raw data without wanting to manage a cluster, this episode is a practical blueprint.

#ServerlessSQL #DataLake #Athena #Presto #AWSGlue #Parquet #ColumnarStorage #Logistics #RealTimeAnalytics #ETL #DataEngineering #Business #Technology #FexingoBusiness #BusinessPodcast #TechnicalCoFounder #StartupStack #Infrastructure

Keep every episode free: buymeacoffee.com/fexingo

This episode explores how a small AI startup replaced cloud-based GPU inference with WebGPU, running neural networks directly in the browser. Lucas and Luna break down the technical details: how WebGPU maps to modern GPUs, the performance trade-offs compared to server-side inference, and why latency-sensitive applications like real-time video filters benefit from client-side compute. They walk through a concrete example—a startup called PixelMind that cut inference latency from 200ms to under 10ms by moving their model to the client. The hosts discuss the challenges: limited memory on mobile GPUs, browser sandbox restrictions, and the need to quantize models without losing accuracy. They also touch on the broader implications for privacy and edge computing. Tune in for a specific, numbers-driven look at one team's journey from cloud to browser.

#WebGPU #MachineLearning #InBrowserML #GPUCompute #EdgeAI #PixelMind #StartupTech #RealTimeInference #ModelQuantization #LatencyOptimization #ClientSideAI #TechDeepDive #BusinessAndTech #FexingoBusiness #BusinessPodcast #Engineering #CTO #TechnicalCoFounder

Keep every episode free: buymeacoffee.com/fexingo

In this episode, Lucas and Luna explore how a growing fintech startup replaced polling-based dashboard refreshes with ASP.NET Core SignalR, cutting server load by 70% and delivering sub-200ms updates to 50,000 concurrent users. They walk through the startup's migration from WebSocket raw sockets to SignalR's automatic transport fallback, the hub architecture that avoided memory leaks, and the one scaling gotcha that nearly broke their Redis backplane. No hype—just the concrete tradeoffs of using persistent connections for live data at mid-scale.

#SignalR #RealTimeDashboards #ASPNetCore #WebSockets #Fintech #SystemDesign #BackendEngineering #Scaling #Redis #ServerPush #LatencyOptimization #EngineeringPodcast #Business #Technology #FexingoBusiness #BusinessPodcast #TheTechnicalCoFounder #StartupEngineering

Keep every episode free: buymeacoffee.com/fexingo

In episode 87 of The Technical Co-Founder Podcast, Lucas and Luna dive into how one startup replaced WebSocket with WebTransport for real-time data streaming. They trace the journey of a live-video analytics startup that switched from WebSocket to WebTransport to cut latency from 150 milliseconds to under 10. Lucas explains the protocol differences — WebTransport runs on QUIC over UDP instead of TCP — and why that matters for applications like live captioning, gaming, and financial tickers. Luna challenges whether the migration was worth it for a small team, and Lucas breaks down the library choices (like Node.js server with webtransport-rs) and the surprising gotchas around TLS certificate requirements and browser support. The episode also covers how the startup used WebTransport's unreliable streams for video frames and reliable streams for metadata, achieving sub-10ms latency. Listeners learn a concrete migration path and the trade-offs between WebSocket and WebTransport for any founder considering real-time architecture.

#WebTransport #WebSocket #QUIC #UDP #RealTimeStreaming #LowLatency #NodeJS #webtransport-rs #LiveVideoAnalytics #BrowserSupport #TLS #UnreliableStreams #HTTP3 #ProtocolMigration #StartupTechStack #EngineeringPodcast #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo

Episode 86 dives into how a small logistics startup, RoutePilot, moved from polling REST endpoints to building a real-time synchronization layer using WebRTC data channels. Co-founder and CTO Maya Chen explains why they chose peer-to-peer data channels over WebSockets for their delivery-driver app, how they handled NAT traversal with a simple STUN server, and the unexpected latency improvements they saw. We walk through the concrete architecture: the signaling server written in Go, the JSON-based message protocol, and how they reconnected after network drops. Maya also shares the one metric that convinced her team to never go back to polling: a 94 percent reduction in server bandwidth costs. If you are building a mobile app that needs live updates without the overhead of a persistent TCP connection, this episode is for you.

#WebRTC #DataChannels #RealTimeSync #StartupArchitecture #RoutePilot #MayaChen #GoLang #STUN #NATTraversal #PeerToPeer #LogisticsTech #MobileEngineering #BandwidthOptimization #WebSocketAlternatives #TechnicalCoFounder #BusinessAndTechnology #FexingoBusiness #BusinessPodcast

Keep every episode free: buymeacoffee.com/fexingo