Optimising CI and Local Dev Setups for Concurrent AI Agents

X postPaulOpinion

After adopting Opus 5.5, commit volume increased fourfold, turning local developer machines and CI pipelines into major bottlenecks. Paul outlines the tooling and infrastructure overhauls used to resolve these constraints.

Key points

  1. A 4x increase in merged commits following Opus 5.5 adoption overwhelmed CI and local 64 GB MacBook Pros running 10+ concurrent agents.
  2. Swapped ESLint and Prettier for oxlint and oxfmt, moving type-aware linting exclusively to CI and cutting file linting time from over 10s to a few seconds.
  3. Adopted TypeScript 7's native LSP over tsserver, halving memory consumption per agent session.
  4. Throttled concurrent resource-intensive tasks to a maximum of three type checks or large test runs simultaneously across worktrees.
  5. Sped up CI setup by ~50x: replaced the standard pnpm/action-setup (which hung for 3-7 mins installing via npm) with a direct binary download from GitHub releases taking ~4s.
  6. Upgraded CI runners to 16 GB to accommodate a 9.5 GB type-checking requirement, reducing check time from 206s (due to paging on 8 GB runners) to 20s.

Why it matters

Running multiple autonomous AI coding agents concurrently shifts productivity bottlenecks from code generation to local machine memory limits and CI pipeline latency.

Analysis

Paul highlights that high-throughput agent workflows quickly expose inefficiencies in standard web development tooling. By migrating to native binaries (oxc, direct pnpm binary downloads, TypeScript 7 native LSP) and matching CI hardware capacity to actual memory footprints, developer workflows can handle high-frequency commits without crashing local environments.

Original post

Paul

@PaulSangleF · 6 October 2026

we merged 4x more commits last week than we did a month ago (thanks opus 5.5). our dev setup and CI couldn't keep up, so we reworked both: • lint: 4x faster • memory per agent session: half • CI setup step: 50x faster • CI type check: 10x faster we often run 10+ agents at t.co/s4aXVw6LXJ

685 likes41 replies1.3k bookmarks75.6k views

View on X
More posts in the thread (2)
  1. @yonasbe Just messaged!
  2. @schiste Most agents are working on completely different features, so it's generally fine.
Selected replies (5)
  • @yonasbe @PaulSangleF Amazing Paul! Had I known you were working on this, I would have sent you our skill :) It basically figures all this out for you by looking at recent runs/PRs and identifying what's actually blocking wall clock time. t.co/hJ3STV95nv
  • @schiste @PaulSangleF How did you manage merge conflicts resolution from the always moving main? Merge queue was enough?
  • @edugiansante @PaulSangleF wow 4x - did the review queue become the next bottleneck?
  • @JacobPetterle @PaulSangleF this is the way
  • @iLegend_AI @PaulSangleF The sky is falling. 4x commits and CI becomes the bottleneck. "CI setup 50x faster" is usually a cache hit, not a speedup. 10+ agents on a 64GB MBP is the real constraint. @grok 50x - cache or actual work removed?

Mentioned links

Summary generated by AI (Gemini 3.8 Flash, 8 October 2026). It may contain errors – the original source is authoritative.

Related entries

Similar in content, found by the AI search.

  • Video

    Video:CURT

    Agentic Development Stack: Harnesses, Models, and Automated Workflows

    Software developer Ross details his updated engineering stack for building agentic AI applications. He explains his use of harnesses like T3 Code, Cursor, and Claude Code, his preferred models Opus 5.5 and GPT-6 Astra, core frameworks including SvelteKit, Convex, and Effect, and an automated Git worktree pipeline featuring Greptail review loops.

    AI & LLMs· Talk

  • X post:Emanuele Di Pietro

    Optimising App Performance with Claude Code and Opus 5.5

    Emanuele Di Pietro shares a prompt for Claude Code that enabled Opus 5.5 to run for three hours, identifying and fixing several overlooked performance bugs in his Electron app Synara.

    3.3kbookmarks181.4kviews

    AI & LLMs· Demo

  • X post:Sanket

    Building a Multi-Agent Development Workflow in Buzz

    Sanket (@tinkerersanky) shares his multi-agent configuration in Buzz, distributing software tasks across six specialised agents. He discusses practices for token management, context preservation, and targeted agent prompting.

    1.3kbookmarks135.9kviews

    AI & LLMs· Opinion

  • Repository:stablyai/orca

    Orca: Orchestrating Parallel AI Coding Agents via Git Worktrees

    Orca is an open-source Agent Development Environment (ADE) enabling developers to run multiple AI coding agents such as ClaudeCode, Codex, OpenCode, or Pi side-by-side across isolated Git worktrees with companion desktop and mobile apps.

    60.9kstarsTypeScript

    AI & LLMs· Tool

Back to the knowledge base

Let's talk about your project

Get in touch

Locations

  • Mattersburg
    Johann Nepomuk Bergerstraße 7/2/14
    7210 Mattersburg, Austria
  • Vienna
    Ungargasse 64-66/3/404
    1030 Wien, Austria

Parts of this content were created with the assistance of AI.