Blog

Updates, insights, and announcements from Charlie Labs.

Posts (41)
Minimal pale geometric building with curved walls, a glowing orange arched entrance, steps, and trees at dusk

How We Built Charlie, Part 15: What We Learned Building Charlie

What we learned about making AI engineering work owned, bounded, verifiable, and useful to a team.
Pale geometric building with three cylindrical towers, a narrow coral-lit passage, palm trees, and a blue sky

How We Built Charlie, Part 14: Making Agent Work Legible

How legible Task state and artifact-linked updates give people one clear next action.
Pale geometric building with tall cream forms, a warm doorway, rounded clouds, and landscaped greenery

How We Built Charlie, Part 13: Proof, Not Vibes

How evidence-backed completion prevents confident reports from outrunning artifacts, provider state, tests, and CI.
Lavender and pink geometric building with orange-lit arches, a reflecting pool, palms, and dense foliage

How We Built Charlie, Part 12: Daemons: Persistent Roles, Bounded Runs

How bounded daemon activations make recurring repository work dependable without a process that runs forever.
Wide photographic sequence of coral and yellow arches casting repeated shadows across a sunlit walkway

How We Built Charlie, Part 11: Context Without Magic Memory

How sourceable, freshness-bound context prevents stale guidance from becoming invisible memory.
Pink geometric building with an illuminated arched doorway, palm trees, and a reflecting pool beneath a pale blue sky

How We Built Charlie, Part 10: Devboxes as the Agent’s Body

How scoped execution environments turn plans into durable, reviewable code artifacts without becoming lifecycle authority.
White geometric house with layered arches, rounded towers, flowering trees, and a deep blue sky

How We Built Charlie, Part 9: Follow-Ups While Work Is Running

How follow-ups change active work safely without starting over or letting stale requirements continue.
A long coral arcade curves beside a reflecting pool toward a cylindrical tower

Your repo sets the rules. Charlie runs the review.

Define repository-owned review policy in Markdown while Charlie selects relevant lanes, verifies evidence, and delivers one coherent pull request review.
Pastel geometric building with glowing orange arches, palm trees, circular forms, and a reflecting pool

Stop prompting your maintenance

A prompt can finish maintenance once. A Daemon owns a recurring, repository-defined job, including when to act, verify, no-op, or escalate.
Geometric concrete building with rounded towers, a coral-lit entrance, palm trees, and a pale blue sky

How to Build a PR Review Agent

Building a PR review agent is easy. Making it reliable means defining authority, completeness, memory, recovery, and a clear operating contract.
A tall cylindrical brick tower and a row of brick arches with yellow and blue interiors against a pale blue sky

I rage-built our email automation. Now we're open sourcing the engine.

Why we are sharing a real email system built with Charlie, and how repo-defined daemons helped keep the work bounded, reviewable, and testable.
Pink brick arcade with yellow and blue archways beneath a pale blue sky

Charlie Credits limits

Charlie enforces the daily and weekly Credits limits included with every plan.
White geometric house with a large central arch, deep blue walls, flowering trees, and stone landscaping

How We Built Charlie, Part 8: Delegation Without Swarms

How explicit delegation keeps parallel work scoped, reviewable, and accountable to one parent Task.
A frontal white geometric building set against a deep blue field with dark rocks and vivid pink flowering plants

How We Built Charlie, Part 7: Task Graph vs Transcript Ledger

How separate Task graphs and transcript ledgers preserve both coordination history and the exact model-visible record.
A layered white architectural complex with cylindrical towers, terraces, trees, and muted yellow-green openings beneath a blue and pink sky

How We Built Charlie, Part 6: The Executor Loop

How Charlie resumes interrupted execution without losing tool order, repeating uncertain effects, or dropping the final result.
Tall white geometric building with rounded towers, softly lit windows, rocks, and trees beneath a pink-blue sky

How We Built Charlie, Part 5: Scheduling Durable Work

How durable Task state keeps work claimable, recoverable, stoppable, and accountable across restarts.
Pastel pink geometric poolside structures with a tiled arch, curved wall, palm trees, and a pale sun

How We Built Charlie, Part 4: Routing Is Ownership

How Charlie prevents duplicate or orphaned work by choosing the right ownership outcome for each Signal.
White geometric building with cylindrical towers, a pale yellow window, and mountains beneath a blue sky

How We Built Charlie, Part 3: Turning Webhooks into Work

How Charlie verifies, normalizes, enriches, filters, and routes provider notifications into trustworthy engineering work.
Pink geometric building with an arched doorway, cactus, and palm trees beside a pale blue pool

A Smarter Charlie, Powered by GPT-5.6

GPT-5.6 brings stronger judgment. An updated Charlie harness puts Sol, Terra, and Luna to work in parallel.
Coral brick architectural forms with a large circular opening framing blue geometric structures, trees, and distant mountains

How We Built Charlie, Part 2: The Task, Not the Chat

A chat can start an engineering request. A durable task keeps it owned while people add context, work is delegated, systems fail, and Charlie reports what happened.
Pale geometric buildings framed by trees and mountains beneath a pastel blue and pink sky

How We Built Charlie, Part 1: Why Dependable Agents Need a Coordination Layer

An agent can make progress in one run and still leave the engineering task unfinished. Dependable systems need a way to keep that task owned across follow-ups, delegation, failures, and evidence.
Muted geometric architectural forms with vertical towers against a cloudy sky

I rage-built email automation because HubSpot wanted me to talk to sales

A SaaS pricing frustration turned into a real production email automation system and a lesson in trusting specs, checks, and repo-owned guardrails.
Warm neutral abstract geometric blog hero artwork with layered technical shapes

The Agent Hangover

Coding agents increase output faster than maintenance loops can keep systems in sync. Daemons give recurring engineering work an owner.
Cool blue abstract geometric background with layered blocks and subtle purple accents

You think you want Loops, what you need is Daemons

Loop Engineering gives you the parts. Daemons give recurring engineering work an owner.
Minimal white building with a glowing orange inset among autumn trees

The Most Important Work Is the Smallest Part

The best software teams save human attention for the decisions that deserve it, use agents to extend that attention, and give recurring operational work an owner.
Abstract muted mountain range beneath a pale sky

The Best Onboarding Flow Is a Pull Request

Start with one useful job, review the setup PR, and expand only after the results are clear.
Abstract dynamic workflow illustration connecting Claude Code and Charlie task orchestration

Claude just discovered workflows. Charlie started there.

Claude just discovered workflows. Charlie started there: durable task-tree orchestration for big migrations, tiny team asks, and everything in between.
Floating white cube-like building with a warm orange interior window

90% cheaper repo inference with gpt-5.4 nano

For bounded orchestration decisions, the right model is often the smallest one that can pass a focused validation loop.
White cylindrical modular building with a warm arched doorway and pale moon

Daemons do the rest — all the necessary work that nobody owns

A taxonomy of recurring Product and Engineering work that doesn't need a human to remember it every week — just a process to hold the role.
Introducing Daemons: Doing the Work That Agents Leave Behind

Introducing Daemons: Doing the Work That Agents Leave Behind

Agents create work. Daemons maintain it. Today we are launching a new product category built for teams dealing with operational drag from agent-created output.
Charlie V2: Introducing the Coding Agent Operating System (CAOS)

Charlie V2: Introducing the Coding Agent Operating System (CAOS)

Charlie V2 is a runtime for durable, multi-step coding work across GitHub, Linear, and Slack. It moves coding agents from one-shot responses to long-running execution that recovers from partial failures and follows through to merge.
White framed building with a central cylinder and warm window panels

Don’t ask if it works. Ask for proof.

AI coding agents will usually answer correctness questions with yes. Ask for proof artifacts instead—outputs, before/after evidence, tests, or explicit reasoning you can inspect.
The End of Local

The End of Local

As agents become autonomous, the local IDE model hits a ceiling — and async remote agents become the default.
Abstract visualization of rising particles, with some rising exponentially higher than others

The Floor is Rising

AI coding agents do not equalize engineering judgment. They amplify it — and the teams that adapt fastest are pulling away.
The task supply problem

The Task Supply Problem

As coding agents get faster, the bottleneck shifts from execution to task generation — and the next productivity unlock is making intent and context as legible as code.
Vibes DIY contributor spotlight featuring CharlieHelps

Link: Vibes DIY spotlights CharlieHelps

Meghan Sinnott interviewed CharlieHelps for Vibes DIY's Contributor Spotlight series — a sharp, funny look at how an autonomous engineer shows up in open source.
Modern white building with pink-lit windows in a mountainous landscape

Why the Next Agent Interface is Shared

Personal agent command centers are a step forward, but the next interface will be shared, multiplayer, and always-on.
Abstract geometric background

For agents, there really are dumb questions

Agents can RTFM. Dumb questions are the ones they could answer by reading the repo—and they cost you time and quality.
Minimal lounge with a circular window looking over a distant landscape

Charlie 2025 - A Recap and What’s Next

We’re really proud of how far Charlie has come in 2025 and wanted to share a brief reflection on the state of agentic software development that Charlie is part of.
Charlie's GPT-5 Upgrade

Charlie's GPT-5 Upgrade

Today, I'm excited to announce a significant update: I've been upgraded to OpenAI's GPT-5.
Large illuminated arched building in a rocky landscape at dusk

The Future of Software Is Agentic

At Charlie labs, we believe the future of software development is agentic, below is our perspective, thoughts, and vision for where things are going.