Skip to content

Knowledge base

Productivity and Workflow

Measured effects of AI-assisted engineering range from slowdowns to large throughput gains. Enablement, measurement discipline and the workflow around the code decide where a team lands, far more than the model or the tool.

Written by

Updated:

Getting Started 2

Our position. We treat AI as a multiplier of what a team already has, so we fix the foundation first and then multiply it (opens in a new tab). Usage is not ROI: we judge a rollout by delivered outcome (opens in a new tab), because a team can use the tool heavily and ship no faster.

Writing code was rarely the bottleneck, so faster generation moves the constraint into review, testing and delivery (Where gains and losses occur). In a randomized trial, experienced developers were about 19% slower with early-2025 AI tools on their own repositories (METR, Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity (opens in a new tab)).

Start here

  • Productivity evidence and quality: What controlled studies and production telemetry show about speed, output, rework, and defects.
  • Measuring impact: How to measure agentic engineering without confusing activity, perception, and business outcomes.
  • Where gains and losses occur: How faster code generation moves constraints into review, testing, delivery, and maintenance.
  • What moves adoption: The organizational conditions that convert access to agentic engineering tools into sustained use.
  • Practitioner practices: The day-to-day sequences practitioners use to research, plan, implement, verify, and learn.

Resources

Research Tools

  • AutoResearch (opens in a new tab) - Research experiment running AI agents that auto-research single-GPU nanochat training, producing reports with citations and structured findings.
  • Dia Browser Skills (opens in a new tab) - Repeatable AI shortcuts for browser workflows in Dia, with a community library to discover and remix.
  • Dovetail (opens in a new tab) - Customer intelligence product that pulls feedback from calls, support tickets, surveys, and reviews, then runs AI analysis, dashboards, and automated insights for product teams.
  • Exa (opens in a new tab) - Neural web-search API that retrieves high-quality, relevant content for AI agents and RAG.
  • Fabric (opens in a new tab) - Practitioner-authored MIT-licensed Go framework that organizes crowdsourced AI prompts (Patterns) into a CLI, REST API, and web interface, running them against 20+ LLM providers.
  • Gluecharm (opens in a new tab) - Requirements and impact analysis tool for software agencies that walks an idea to IDE-ready specs across feature, experience, data model, service, and tech stack.
  • Manus (Research AI) (opens in a new tab) - AI agent that executes tasks and automates workflows beyond just answering questions.
  • Parallel AI Search (opens in a new tab) - Web search infrastructure built for AI systems with Search, Deep Research, and Task APIs plus a Dataset Builder, returning evidence-backed and cross-referenced outputs.
  • Perplexity Comet (opens in a new tab) - Perplexity’s AI browser that runs research and search inside a native browsing experience.
  • Promptmetheus (opens in a new tab) - Prompt engineering IDE for composing, versioning, A/B testing, and auto-evaluating prompts across 100+ models.
  • RepoMirror (opens in a new tab) - Experimental tool that ports repositories between languages and frameworks by running coding agents in continuous loops.
  • Sourcebot (opens in a new tab) - Self-hosted code search across GitHub, GitLab, and other forges with regex search, goto-definition, find-references, and natural-language Q&A over your repos.
  • pi-autoresearch (opens in a new tab) - Autonomous experiment loop for the pi agent that proposes ideas, measures results, and iterates on any metric by separating domain-agnostic infrastructure from domain skills.
  • witr (opens in a new tab) - “Why is this running?” Apache 2.0 Go CLI/TUI that explains the causality behind a process, service, or port, replacing manual correlation across many CLI tools.

Communication & Collaboration

Engineering Intelligence & Analytics

  • Bloomfilter (opens in a new tab) - Process intelligence platform giving engineering leaders visibility, traceability, and governance over how humans and AI agents execute SDLC work.
  • Linear (opens in a new tab) - Product and issue tracker repositioning around AI agents, adding native agents, reusable skills, and triage automations to its workflow.
  • PostHog (opens in a new tab) - Open-source product analytics suite bundling analytics, session replay, feature flags, and a data warehouse under usage-based pricing.

Terminals & Editor Setups

  • Alacritty (opens in a new tab) - GPU-accelerated, cross-platform OpenGL terminal emulator focused on speed and a minimal config surface.
  • LazyVim (opens in a new tab) - Neovim distribution on top of lazy.nvim that ships an opinionated IDE setup with curated plugins and sane defaults.
  • Warp (opens in a new tab) - Agentic development environment built around the terminal, with built-in AI coding agents.
  • WezTerm (opens in a new tab) - Cross-platform terminal emulator and multiplexer in Rust with panes, tabs, ligatures, color emoji, and dynamic color schemes.

Personal Productivity Tools

  • Tolaria (opens in a new tab) - Open-source local-first note app storing notes as Markdown files with wikilinks, Git versioning, and direct AI providers, a second brain for the AI era.

Fun

In this section

  • What moves adoption

    Sustained adoption follows dedicated learning time, clear policies, peer mechanisms such as champions and sharing channels, and role-specific workflows packaged for reuse. Top-down usage targets produce compliance theater, and access alone does not produce use.

  • Productivity evidence and quality

    Measured productivity gains from AI coding are modest and conditional, and one randomized trial found a slowdown: they vary by task, codebase and enablement, self-reports overstate them, and higher output raises churn, defects and review load unless quality gates scale with it.

  • Measuring impact

    Measure agentic engineering by pairing AI metrics such as adoption, AI-assisted share and spend with the delivery metrics that predate AI, such as throughput and change failure rate. AI metrics show what is happening. Core metrics show whether it works.

  • Practitioner practices

    As of October 2026, practitioners hand agents whole tasks with the tools to see the delivery cycle and ask for proof instead of steering each step: the plan goes into a committed file, failed attempts are discarded instead of patched, questions state the goal rather than the method, and exact repeated operations go into scripts.

  • Where gains and losses occur

    AI speeds up writing code, which was rarely the bottleneck. The constraint moves into review, testing, change approval and operations, so system-level gains come from fixing coordination and handoffs, not from generating code faster.