September 2026 Summaries

Large language models have shifted much of software development toward prompting agents to handle complex tasks, leading some experienced developers to rely less on traditional IDEs, but human judgment remains important for directing results and reviewing generated code. The author argues that agent-produced code often requires cleanup because it can be overly defensive or prematurely abstracted, particularly in familiar areas such as frontend work. IDEs remain valuable for focused editing, code navigation, definition lookup, and tracing logic, capabilities refined over decades that many agent-centric tools have not yet matched. Referencing Jason Torres’s Commit Your Code Conference 2026 talk, the piece suggests that IDEs may evolve into collaborative spaces where developers and agents work together rather than replacing one another. While the author is increasingly using agent-oriented development environments such as Kepler, T3 Code, and Opencode Desktop, they want these tools to offer a stronger VS Code-like experience for reviewing and refining generated code.

GitLens 19.2 introduces Scope to Worktree, which re-centers the existing Commit Graph on a selected Git worktree so its working changes, status, ahead/behind indicators, and Git actions all reflect that worktree without reloading history or losing graph state such as scrolling, selections, searches, and filters. Users can scope worktrees through the sidebar, WIP rows, or Commit Graph commands, with visual indicators showing the active worktree, while separate settings control how scoping interacts with Focus, which instead narrows visible history to a branch. The release also adds the ability to resume Codex, GitHub Copilot CLI, and OpenCode agent sessions in either a terminal or their respective extensions, with configurable default resume targets. Other changes include saving Commit Graph layouts as workspace defaults, running a default VS Code task for any worktree, customizing GitLens keyboard shortcuts, retiring the Home view, and adopting the VS Code display language with experimental Spanish and Chinese translations.

GitKraken presents a pull-request review workflow designed to help developers focus on meaningful code changes while retaining human control over approval decisions. Reviewers begin in Launchpad by identifying a PR’s scope and context, use the Code Review diff viewer to filter formatting noise and inspect logic changes, then consult an AI panel for summaries, questions, mergeability, conflicts, reviews, and CI/CD status. After evaluating AI-generated findings, reviewers leave comments, request changes, manage participants, or approve the PR directly through its GitHub-connected workflow. The company promotes Kepler for organizing agent tasks, GitKraken Desktop for repository and conflict management, and GitLens for IDE-based history and authorship context, alongside related cloud collaboration features and a limited-time 60% discount on GitKraken Pro.

Kepler is GitKraken’s agentic development environment, designed to help developers coordinate multiple AI coding agents working simultaneously across repositories rather than merely generate code faster. It integrates with existing agents, models, issue trackers, and Git hosts, including Claude Code, Codex, GitHub Copilot, Cursor, Jira, Linear, GitHub Issues, and GitLab Issues, while using isolated worktrees to prevent parallel agent sessions from conflicting. Its Agent Graph provides live visibility into tasks, sessions, tool calls, and stalled work, while configurable Actions supply agents with relevant issue, repository, and diff context. Kepler also manages the workflow beyond an agent’s completed task by supporting review, staging, commits, pull requests, and merges, emphasizing that agent output is not production-ready until it is visible, reviewed, and integrated. Available for Windows, macOS, and Linux, it supports remote and WSL environments and aims to fit into teams’ existing development toolchains.

Enterprise pull request review tools aim to reduce review delays, improve code quality, and provide governance across large, distributed development organizations. The comparison evaluates GitKraken, LinearB, CodeScene, Code Climate, Jellyfish, and Waydev based on workflow control, collaboration capabilities, enterprise security, Git-host compatibility, engineering analytics, and AI assistance. GitKraken is presented as an integrated review platform with inline code suggestions, AI-supported diff summaries, pre-PR collaboration, centralized dashboards, DORA metrics, and enterprise controls, while LinearB and Waydev emphasize workflow and merge-quality analytics, CodeScene focuses on behavioral code health and refactoring recommendations, Code Climate provides static quality scoring and merge gates, and Jellyfish connects review activity to broader business and delivery reporting. Across the tools, a central distinction is between platforms that support hands-on review actions and those primarily designed for reporting, automated checks, or management visibility. AI-assisted reviews can help teams filter formatting noise and prioritize meaningful logic changes, while enterprise governance requires access controls, auditability, approval policies, and visibility into review bottlenecks, cycle times, and workload distribution.

Engineering intelligence platforms help teams assess software delivery by combining data from Git repositories, CI/CD systems, project trackers, and other tools, with key evaluation factors including coverage of the four DORA metrics, AI coding-tool impact, code quality signals, integrations, setup effort, and developer-focused reporting. The comparison identifies GitKraken Insights as a platform combining DORA metrics, AI adoption and agent analysis, code quality monitoring, surveys, and rapid setup, while noting that some advanced functions and historical data depend on plan level. DX emphasizes survey-based developer experience measurement and benchmarking, Swarmia focuses on Git and project-tracker workflow visibility and team agreements, and LinearB pairs delivery metrics with pull-request automation. Faros AI is positioned around aggregating data from more than 40 engineering tools into customizable dashboards, including newer AI adoption tracking, whereas Allstacks specializes in forecasting delivery dates and project risks using historical workflow data. The text argues that organizations adopting AI coding tools should compare delivery performance, stability, rework, and defects before and after adoption, using DORA metrics alongside AI-specific and code-quality data to make evidence-based decisions.

Kepler can now use the minimalist Pi agent through the third-party pi-acp proxy, which implements the Agent Client Protocol used by agent development environments to communicate with CLI harnesses such as Claude Code, Codex, and OpenCode. While setup is straightforward after installing Pi and pi-acp, Kepler currently displays inaccurate permission modes and thinking levels because pi-acp hardcodes effort settings rather than using Pi’s newer available-thinking-levels function, and because Pi’s intentionally permissive default behavior does not map cleanly to Kepler’s expected permission-level system. The author plans to contribute fixes to pi-acp, while Kepler’s team is considering how to support agents without permission modes. Although Pi’s plugin ecosystem can add features such as subagents and research tools, the author argues that ACP support is fundamental enough to be integrated directly into Pi rather than maintained through a plugin or external proxy.

GitKraken support highlights four current topics: Microsoft plans to deprecate some personal access token configurations and modify OAuth in early December, and GitKraken says it will implement a solution before existing functionality is affected, though details are pending Microsoft documentation. GKAI subscriptions use individual weekly credit pools first, then organization-wide shared pools for continued AI use and automations, with purchasable add-on credits available if shared credits are exhausted. Kepler users on Pro plans and above can connect multiple AI, cloud, and Git service accounts through Kepler settings and gitkraken.dev, then refresh integrations to synchronize accounts, organizations, and repositories. For GitKraken Desktop performance or display problems on Linux or WSL, disabling GPU acceleration with the `gitkraken --disable-gpu` command generally resolves lag, sizing, and drag-and-drop issues.

Internal hackathons give teams space to explore creative ideas, address neglected backlog items, experiment outside normal responsibilities, and strengthen collaboration. The author describes a non-coding Fairwinds hackathon project involving a Kubernetes-themed parody song, while citing PostHog’s use of hackathon prototypes, including Session Replay, as examples of ideas that can become successful production features without lengthy planning processes. GitKraken similarly originated during an Axosoft hackathon when employees identified an opportunity to improve the developer experience of working with Git. Although GitKraken’s Kepler product was not itself created at a hackathon, earlier projects for monitoring AI coding agents and connecting assistant conversations with Git history helped validate its underlying concept. The account argues that hackathons are valuable not only for major corporations but also for smaller companies and startups, where they may produce meaningful products or strategic changes.

Engineering teams evaluating platforms for parallel AI coding agents should assess agent compatibility, automatic Git worktree isolation, multi-repository coordination, real-time monitoring, integration with issue trackers and Git hosts, traceable code review workflows, and controls to pause or redirect flawed agent sessions. Deployment options also matter, particularly for organizations that require local, hybrid, or self-hosted execution to meet security policies. Teams should measure effects on delivery through metrics such as deployment frequency, lead time, and rework, while ensuring the platform supports appropriate governance, access controls, and audit requirements as usage scales. The piece presents GitKraken Kepler as an example of an agent-agnostic orchestration environment that integrates these capabilities, arguing that simpler CLI tools may suit individual experimentation while larger teams working across repositories need more centralized coordination and visibility.

Pull request bottlenecks often result from unclear reviewer ownership, fragmented repository views, noisy diffs, late feedback, merge conflicts, and limited visibility into review timing. The piece argues that effective code review platforms address these issues through centralized dashboards, AI-assisted diff filtering and summaries, early sharing of uncommitted work through cloud patches, automated reviewer assignment and escalation rules, and tools for suggestions beyond changed lines. It presents GitKraken’s Launchpad, Code Review, GitLens Cloud Patches, Automations, Desktop, and Insights as an integrated set of features for tracking pull requests, identifying risky changes, preventing conflicts, measuring pickup and merge times, and maintaining consistent workflows across desktop, IDE, CLI, and browser environments. It also emphasizes enterprise access controls and integrations with Git hosting services, issue trackers, and CI/CD systems, framing centralized context, automation, and measurable workflow data as priorities for teams seeking faster reviews without reducing quality.

Engineering intelligence platforms aggregate data from tools such as Git hosts, CI/CD systems, and issue trackers to assess software delivery performance, code quality, developer experience, and increasingly the effects of AI coding tools. The comparison evaluates six alternatives to Jellyfish—GitKraken Insights, LinearB, Swarmia, DX, Allstacks, and Faros AI—against DORA metric coverage, AI impact measurement, setup speed, integrations, leadership reporting, and code-quality visibility. GitKraken Insights is presented as a fast-to-deploy platform combining DORA metrics, AI ROI analysis, pull request flow data, code-quality indicators, and developer surveys; LinearB emphasizes workflow automation and Git analytics; Swarmia focuses on transparent developer-facing metrics and investment allocation; DX combines delivery data with survey-based developer experience research; Allstacks specializes in AI-assisted delivery risk forecasting and natural-language data exploration; and Faros AI offers highly customizable, open-source-oriented analytics for data-savvy teams. The discussion argues that organizations replacing Jellyfish should consider whether they need only delivery metrics or a broader view that links delivery speed with reliability, code maintainability, developer sentiment, costs, capacity, and the post-merge outcomes of AI-assisted code.

Jim Shaw announces his appointment as GitKraken CEO, succeeding Matt Johnston after five years of company growth. Shaw says he joined because GitKraken addresses the challenge of helping software teams use AI-driven development capacity effectively while preserving context, quality, and control across multiple agents, models, and repositories. He highlights GitKraken Insights for measuring AI’s business and engineering impact and Kepler, an agentic development environment that enables developers to use multiple coding agents and preferred models across repositories. Shaw emphasizes the company’s commitment to developer choice rather than closed AI ecosystems, drawing on his systems-oriented experience in automotive and 15 years in leadership roles at Acquia, while pledging to learn how AI is creating both value and friction for customers.

AI code-to-merge platforms are presented as tools for managing the expanding workflow created by AI coding agents, covering planning, code generation, pull request review, feedback resolution, and merging rather than focusing on a single development stage. Key evaluation factors include support for multiple interchangeable agents, real-time monitoring of agent activity, deep integration with Git hosts and issue trackers, isolated workspaces for parallel tasks, tools for reviewing and committing agent output, reusable task templates, security and data-location controls, production-outcome measurement, and rapid onboarding. The discussion uses GitKraken Kepler as an example, describing its support for agents such as Claude Code and Codex, integrations with major Git and project-management systems, dedicated worktrees, local execution options, and analytics through GitKraken Insights. It argues that platforms with broad lifecycle coverage and flexible integrations can reduce context loss, manual handoffs, merge conflicts, and vendor lock-in as AI-generated code volume and pull request sizes increase.

Distributed software teams often face code review delays caused by asynchronous handoffs, missing rationale for changes, poorly matched reviewer assignments, late feedback, formatting-heavy diffs, unexpected merge conflicts, limited performance metrics, and knowledge concentrated in a few reviewers. The piece argues that conventional review platforms tend to focus on displaying diffs while neglecting these broader workflow issues, which can increase cycle time and obscure risks. It presents GitKraken’s DevEx Platform as an integrated alternative, citing features such as centralized views of PRs and tasks, linked issue context, AI-generated summaries, workload- and expertise-aware reviewer routing, early sharing of work in progress, noise-reduced diffs, conflict-resolution support, and delivery analytics. It recommends that teams audit their review process, identify the most costly bottlenecks, track measures such as pickup time and review duration, and rotate reviewers to improve collective ownership and reduce dependency on individual experts.

A webinar poll of developers found that while AI-agent adoption is widespread, only 24% can demonstrate with data that agents are delivering value, leaving most reliant on intuition, uncertainty, or concerns about increased spending. Developers’ primary measurement priority is whether agents operate efficiently without requiring extensive supervision, rather than simply tracking token or subscription costs. Many teams already use multiple agents simultaneously, but hidden costs such as rework, code review, and merge delays nearly equal visible spending concerns. GitKraken positions its Kepler agentic development environment and Insights reporting product as tools to coordinate multi-agent workflows, manage commits and conflicts, preserve human oversight, and measure the efficiency and downstream impact of AI-assisted development.

Kepler, GitKraken’s agentic development environment in public preview for Windows, Mac, and Linux, is designed to help developers coordinate multiple coding-agent sessions across repositories, issues, and pull requests. Its tutorial series explains how tasks serve as containers for agent sessions, changes, and related resources, and can be started from issue trackers, pull requests, or blank exploratory prompts. The platform supports any agent or model, including side-by-side use of tools such as Claude Code and Codex, while Actions provide configurable shortcuts for planning, reviewing, addressing feedback, and launching custom prompts with attached context. Kepler’s workspace includes task lists, Kanban views, session tools, and project context, while its Agent Graph provides live visual tracking of agent activity, permission requests, errors, tool calls, and sub-agent work across individual tasks or all active work.