Agentic Development
Building software with AI: coding agents, agent frameworks and tooling, developer workflows, software engineering practice, how engineering teams change with AI.
Delivery Hero's coding agent merges 1,800 pull requests in Q1
Delivery Hero's internal system, HeroGen, exceeded its quarterly target by 18 times, merging 1,800 production-bound pull requests. The company is centralizing its AI stack to manage costs and governance across 1,500 engineers.
G2 defines trust layer framework for probabilistic AI agents
G2’s CPO Alexis Zhang outlined a framework for verifying AI agents, emphasizing verification, guardrails, and observability. The company is also deploying 'AI interns' to accelerate engineering workflows.
Plio reports 90% daily AI adoption and launches agentic platform
Plio’s engineers report 90% daily use of AI coding tools, following a cross-functional hackathon that aligned strategy. The company is developing an agentic CRM platform to enable natural language analytics for marketing teams.
Agentic coding tools disrupt traditional software engineering social contracts
Rafi argues that agentic coding has increased individual power but broken collaboration norms, leading to low-quality open-source projects. He predicts a rise in demand for engineering and product managers to redefine these contracts.
Svix launches Poke spec as Standard Webhooks gains major adoption
Svix introduced the Poke specification to support agent-driven event integration, noting that traditional push webhooks are ill-suited for ephemeral environments. The Standard Webhooks spec has been adopted by OpenAI, Anthropic, and Google Gemini.
Agentic coding risks increasing technical debt and system complexity
Svix CEO Tom Peck-Owen warns that rapid AI code generation leads to unsustainable complexity, akin to offshoring. He argues that intentional simplicity and deterministic checks are required to manage the resulting technical debt.
Uber's Minion agent generates 15% of production pull requests
Uber's background coding agent now handles a significant share of its engineering output, relying on pre-existing platform infrastructure to validate code. The company argues that mature developer environments are essential for scaling autonomous software development.
Blind test compares human and AI architecture decision records
A keynote experiment compared architecture decision records written by a human and an AI for a bicycle shop system. Audience voting showed mixed preferences, with humans favored for authentication and availability, while the AI won on frontend technology.
Tesla defines software factory architecture with tiered feedback loops and human roles
Tesla has outlined a 'software factory' model that combines repeatable workflows with three distinct feedback loops to manage autonomous AI development. The company assigns specific human roles to oversee this system, ensuring that automation enhances rather than replaces engineering oversight.
Linear B adopts spec-driven development to accelerate AI engineering workflows
Linear B has transitioned its engineering team from ad hoc coding to a mandatory spec-driven workflow to manage AI-generated code. The methodology reduced the timeline for a complex feature from six months to less than one month.
Vention framework shifts engineering focus to spec-driven development
Vention's 'AI-First-SDLC' framework prioritizes spec-driven development to minimize manual effort in converting requirements to code. The company argues that traditional metrics are insufficient for measuring AI impact in engineering teams.
Experts warn one-hour workshops insufficient for product discovery skills
Product experts argue that short training sessions cannot effectively teach continuous product discovery, which requires long-term practice of distinct habits. They recommend longer, structured programs to address organizational barriers before implementing training.