Atlassian engineers Andy Wong, Andrei Bocan, and Shane Wolf presented a workshop on assembling an AI-driven software development lifecycle around Jira's coding agent, AI review, and planning tools. A live demo shows an engineering standard encoded as a reusable code review rule applied across selected repositories. The session matters to practitioners because it frames agents as fitting into existing team process — Jira items, organizational context, and human review — rather than replacing the SDLC.
Nnenna Ndukwe of Qodo AI demonstrates how to build quality gates into agentic coding workflows using SignalPay, a deliberately narrow FastAPI payments application. A refund feature passes through planning, implementation, local checks, and pull-request review, with Qodo review findings tied to codebase and organizational rules. The workshop's reusable artifact is a structured process keeping plans, code changes, and review evidence connected.
Nnenna Ndukwe of Qodo AI demonstrates how to build quality gates into agentic coding workflows using SignalPay, a deliberately narrow FastAPI payments application. A refund feature passes through planning, implementation, local checks, and pull-request review, with Qodo review findings tied to codebase and organizational rules. The workshop's reusable artifact is a structured process keeping plans, code changes, and review evidence connected.
Serokell turned its internal engineering guidelines into 15 Claude Code skills packaged as a public plugin. The skills teach coding agents team-specific conventions like repo bootstrapping, testing expectations, PR structure, CI checks, and code review. It solves the problem that agents know general tools but not a particular team's practices.
Serokell turned its internal engineering guidelines into 15 Claude Code skills packaged as a public plugin. The skills teach coding agents team-specific conventions like repo bootstrapping, testing expectations, PR structure, CI checks, and code review. It solves the problem that agents know general tools but not a particular team's practices.
Anthropic published a playbook for AI-native software development, outlining six stages from plan to maintain, each ending in a committed markdown artifact. The playbook argues that line-by-line code review becomes impractical when agents write most of the code, replacing it with artifact-based gates and human approval checkpoints.
Anthropic published a playbook for AI-native software development, outlining six stages from plan to maintain, each ending in a committed markdown artifact. The playbook argues that line-by-line code review becomes impractical when agents write most of the code, replacing it with artifact-based gates and human approval checkpoints.
Orbit is Flywire's internal platform for placing model steps within deterministic engineering workflows, such as code review, where the model reads code but never modifies it. Workflows are defined in YAML, reviewed via merge requests like any other code, and run on existing deployment infrastructure. It solves the problem of using LLM judgment predictably in engineering processes without giving up human review or control.
Orbit is Flywire's internal platform for placing model steps within deterministic engineering workflows, such as code review, where the model reads code but never modifies it. Workflows are defined in YAML, reviewed via merge requests like any other code, and run on existing deployment infrastructure. It solves the problem of using LLM judgment predictably in engineering processes without giving up human review or control.
Atlassian cofounder Mike Cannon-Brookes told Fortune that accountability for AI agent actions rests with the end user, not the model. On Oct. 6 Atlassian rolled out its Agentic Multiplayer Protocol, a control plane that gives agents distinct profiles and owner attribution across its platforms.
Atlassian cofounder Mike Cannon-Brookes told Fortune that accountability for AI agent actions rests with the end user, not the model. On Oct. 6 Atlassian rolled out its Agentic Multiplayer Protocol, a control plane that gives agents distinct profiles and owner attribution across its platforms.