Stack Overflow engineering blog

Loading
The pioneering coding tricks that built Doom and Quake
In a podcast recorded at WeAreDevelopers 2026, John Romero, creator of Doom and the WAD file format, discusses the self-modifying code and blitting techniques that let Wolfenstein run quickly on early PCs. He also covers the mod-friendly architecture behind Doom's success and the creative problem-solving behind Quake, arguing that human ideas outperform AI coding agents.
Agentic search doesn't find pages, it assembles answers
Stack Overflow podcast episode in which host Ryan interviews EXA co-founder Jeff Wang about search infrastructure for AI agents, covering the scale needed for trillions of agentic queries, custom embeddings, vector databases, and deduplicating redundant web content for trustworthy answers. The readable body contains only the episode blurb and sponsor promotion, so the technical discussion itself could not be assessed.
Your coding agent shouldn't grade its own homework
Momentic's engineering team argues that coding agents should not verify their own work, citing an audit where only 9.6% of 1,000 AI-authored unit tests were useful and a comparison showing agents that wrote tests did no better than those that did not. They describe an independent QA agent that tests live applications from the original requirements, which their benchmark found surfaced 3-5 times as many bugs as agents checking their own output.
Customer support is the gnarliest AI problem
A podcast conversation between Ryan and Airbnb CTO Ahmad Al-Dahle covering multi-party AI agents for customer support. It touches on using rigorous evaluations to balance latency against accuracy and on the higher judgment AI coding tools require from engineering teams. The available page text gives only the high-level topics, not the specifics of the discussion.
Never Share Your Password Again — The OAuth 2.0 Solution
The post explains how OAuth 2.0 lets a third-party app receive scoped, time-limited, revocable access without ever handling the user's password. It walks through the Authorization Code flow with PKCE using concrete example requests and responses. It also details how the state, PKCE, redirect_uri, and short-lived code parameters each defend against specific attacks, and why scopes limit the damage from a breach.
Welcome to life in the chip age
A podcast-style interview in which Ryan speaks with Rakesh Kumar, a University of Illinois computer engineering professor, about the complexity of semiconductor manufacturing, the environmental and geopolitical footprint of AI data centers, and how AI is changing chip design. The page body contains only a teaser and a book plug, so the discussion itself could not be assessed.
Don’t Build the Reusable Component Yet
The author describes a frontend case where merging three similar item workflows into one component produced a growing set of mode and variant flags, coupling unrelated decisions and slowing changes. The post proposes a practical test for reuse: abstract only when copies share the same reason for change, such as shared date formatting, design tokens, accessibility behavior, or standards-defined shapes. It also outlines how to undo a premature abstraction by inlining and re-splitting the code.
Your .NET parallel loop can become serial in a one-CPU container
The author shows that .NET Parallel.ForEachAsync defaults to a concurrency limit based on ProcessorCount, which can serialize an I/O-heavy loop in a one-CPU container and double processing time. Using simulated lab measurements, the post compares four strategies and argues for separate, explicitly configured concurrency budgets per downstream dependency, with separate permit-wait and call timeouts. It shows that a single shared limit can overload a fragile system of record.
Oh, your agents need OAuth2, too
A podcast conversation recorded at LDX3 in which Dan Moore of FusionAuth discusses adapting established OAuth2 protocols for autonomous AI agents. The episode also covers why agents need distinct identities for security and how AI coding tools are shifting developer focus toward core business problems. The page body offers only a high-level summary, with no concrete protocol details or implementation specifics.
It’s ok that computers are (sometimes) better than us
A Stack Overflow podcast conversation between host Ryan and Honeycomb CTO Charity Majors on applying immutable infrastructure ideas to software development, treating code like cattle rather than pets. The episode summary also touches on revising system specifications for AI-generated code and the ethical limits of observability versus human surveillance. Only the episode summary was available, so the depth of the discussion cannot be assessed.
The reports of Stack Overflow's death are greatly exaggerated
Stack Overflow's CEO-level communications responds to predictions of the site's decline, arguing that AI has changed how developers access knowledge without eliminating the underlying knowledge or the company. The post cites a 2026 developer survey and highlights new products, including Stack Overflow for Agents, Developer Story, and Stack Internal. It contains little technical detail and mainly promotes the company's strategy.
A JSON schema Is not an AI governance layer
The author argues that JSON Schema validates only the shape of agent-action payloads, not the semantic rules that determine authorization. Using a parser-differential bug found in the PIC protocol, the post shows how duplicate identifiers, normalization choices, and self-asserted trust create divergent verdicts across implementations. It recommends separating schema validation, semantic verification, and policy evaluation, and using shared conformance vectors to pin down ambiguous behavior.
Delivering controlled and concentrated context
Stack Overflow announces its October release for Stack Internal and Stack Internal Community, including Scopes for permission-governed knowledge spaces, a GitHub connector, service accounts for OAuth-based app access, a validation API, and expanded moderation controls. The post is a feature overview with little implementation or architectural detail.
Elevators, toasters, and the birth of Java
A Stack Overflow interview in which host Ryan talks with James Gosling, creator of Java at Sun Microsystems, about the early networking problems that drove interconnectedness, the reliability demands of consumer electronics makers that shaped the Java virtual machine, and the evolution of garbage collectors for large-scale concurrent memory management. The page body contains only the summary and site boilerplate, so this description relies on the feed summary.
Rumors of the SaaS demise are greatly exaggerated
This podcast episode, recorded at LDX3 by LeadDev, features engineering leader Nivia Henry discussing the supposed SaaS apocalypse and the build-versus-buy decision in the era of AI. The conversation also covers understanding system architectures to manage long-term infrastructure costs. The page provides only a summary and speaker bio, so no specific techniques, data, or trade-offs are visible.
Is AI really reshaping the developer market?
A Stack Overflow blog post, drawing largely on Lemon.io hiring data, that examines how AI has reshaped developer demand by seniority, the shift toward contractors and independent senior engineers, and which roles take longest to fill. It offers general hiring advice such as assessing judgment over tool lists and reducing internal process delays. The content is mostly labor-market statistics rather than engineering practice.
The Regression Tests Nobody Writes: Capturing How Software Actually Behaves
The post argues that regression suites usually capture developers' intentions rather than actual system behavior, and that undocumented quirks often break downstream users. It explains characterization tests, golden master and snapshot testing, and recording real service traffic at the network level to pin behavior, including side effects like database calls. It also covers how to handle noisy fields and where this approach helps most, such as legacy refactors, framework upgrades, rewrites, and...
Building Immersive Virtual Economies: An Introduction to KursarScript (KSPL)
The post introduces KursarScript (KSPL), a language described as built for managing virtual currencies, virtual assets, and transactions in VR environments. It translates the language's core concepts, including VirtualItem, VirtuCard, and VirtualTerminal, into basic Python classes. The examples are simple and the overview stays at the surface, offering little technical depth.
Survey Says: You'll swap the model, but everything you build around it is staying put
Stacklok surveyed 520 North American technical leaders running LLMs or agents to see how concerned they are about lock-in to frontier labs. The survey finds most organizations already use multiple model providers, while integrations, multi-step agent workflows, and guardrail config are the hardest parts to migrate. The post closes with practical advice on keeping tool definitions, orchestration, and evals portable across model endpoints.