Blog
Plain software, written down.
Notes on speed as respect, bringing your own agent, and refusing the AI-era pivot. From the team building Radial.
Jira went AI-native. Developers asked for the opposite.
Atlassian shipped an AI-native Jira where every agent action is governed, measured, and tied to a business outcome. The top comment was not gratitude. Here is what developers actually want, and the pledge that says it out loud.
Radial vs Shortcut: same BYO-agent thesis, a price that never moves
An honest guide to Shortcut alternatives in 2026. Where Shortcut is genuinely strong, why its agent story is already good, and the one thing a tracker building its own AI PM cannot promise you: a flat locked price and a binding no-meter pledge.
The anti-roadmap: the features we will never ship
Most roadmaps are a list of what is coming. This is the opposite: the concrete list of things Radial has promised never to build, why each one is on the list, and what happens to your bill if we break the promise.
Running agents in parallel? They need one shared record to coordinate around
Parallel agents split a job into concurrent subtasks. The hard part is not launching them, it is keeping them from stepping on each other. That is a shared-state problem, and the state your agents fight over most is the task list. Here is where it should live.
Shipped this week: deeper views, OAuth apps, webhooks, and email notifications
A week of boring-on-purpose shipping: Linear-grade view depth, third-party OAuth apps with confidential clients, HTML prototype Previews, HMAC-signed outbound webhooks, and email notifications with real preferences.
Your background agent needs somewhere to pick up the task and file the result
A background agent is defined by two touch-points on your tracker: it picks up a task, and it files the result. Here is where that record should live so the agent can read and write it over an API.
Why the results appear before the spinner would: how a tracker feels instant
The trick to a fast issue tracker is not a faster server. It is keeping the working set in the client so filtering, sorting, and editing happen locally, and the network catches up in the background. Here is the architecture, and where the honest limits are.
GitHub stores the code, not the decision. Where the "why" should actually live.
An architecture decision record captures one significant technical choice: its context, the decision, and the consequences. Here is what an ADR is, when to write one, and why the decisions that matter most are the ones that never reach a doc/adr folder.
Agentic coding ships the 80%. Your tracker holds the other 20%.
Agentic coding is when an AI agent plans, writes, and tests code with little human input. It ships the functional 80% fast. The durable 20%, the decisions and the record, has to live somewhere. Here is where.
You don't want a git issue tracker. You want issues that live next to your code.
Devs keep reinventing the git-native issue tracker: a .issues folder of Markdown committed to the repo. The instinct is right. Here is the fast, scriptable version that links branches, PRs, and commits to issues without asking you to hand-roll it.
Your Claude Code sub-agents keep losing the thread. Give them one shared record.
Claude Code agents and agent teams are real now, but the moment you run more than one, they lose track of what the others did. Here is where the shared record should live.
Graduating from Trello: when a board needs to become a system
An honest guide to Trello alternatives in 2026. Where Trello still fits, why engineering teams outgrow the board, and the case for a fast issue tracker at one flat locked price, agents ride free.
An issue tracker your agents can actually drive
Local trackers like Beads give one coding agent a memory. When a team runs several agents and humans need to see the same work, the record has to be hosted, shared, and driveable over MCP and a CLI. Here is what that looks like.
File issues from CI with the Radial CLI and --json
Turn a failed CI job into a filed issue automatically. Install the radial CLI, run radial create with --json, and pipe the returned RAD-123 identifier back into your pipeline: no dashboard, no seat cost.
The best AI for project management isn't in the tool. It's your agent.
Every tracker is bolting on an AI copilot and a credit meter. The better AI for project management is the agent you already run, driving a fast tracker through real interfaces, for free.
The agile project management tool that is just a fast issue tracker
Most agile project management tools are everything-suites with a board bolted on. Here is the case for the smaller thing: a fast, scriptable issue tracker with real sprints, and nothing you have to grow into.
monday makes you pre-buy AI credits. We sell one flat number.
An honest guide to monday.com alternatives in 2026: what monday is good at, when to leave, and why a dev team that just wants a fast tracker should skip the AI credit meter for one flat locked price, agents ride free.
Scoped API keys: one key per agent, read-only, triage-only, or full
Give each agent its own Radial API key with exactly the access it needs: read-only, triage-only, or full write. Least privilege for an agent fleet, plus how to mint, scope, and revoke keys from the CLI.
Will your tracker still exist next year? Switching-cost anxiety, and the one-command exit
The quiet reason you stall on adopting a tracker isn't the setup. It's the fear of being stuck when it changes, gets acquired, or shuts down. Here is how to price that risk to near zero.
A fast board and real sprints, without the scrum tax
A scrum board is a two-dimensional view of one sprint's work: columns for status, cards for issues. Here is the version that keeps the board and the sprint and drops the ceremony, plus how to drive it from the terminal.
ClickUp wants to be everything. Radial does one thing.
An honest guide to ClickUp alternatives in 2026. Where the everything-app fits, why engineering teams outgrow it, and the case for a fast tracker that does one job at one flat locked price, no AI meter.
Project management for startups, without the bloat
The best project management tool for a startup is the fastest one that does one thing well. Here is the lean version: a keyboard-first tracker you script from the terminal, one flat locked price that does not climb as you grow, and agents that ride free.
Engineering project management without the overhead
Engineering project management is planning, tracking, and shipping software work. Most tools bury that under process. Here is the lean version: a fast tracker your team runs the work in, scriptable from the terminal, with one flat locked price and no AI meter.
Migrating from Linear to Radial in one command
How to move off Linear to Radial: export your workspace, dry-run radial import --from linear, then bring issues, projects, labels, comments, and history across in a single run. Plus what changes, and what stays the same.
Migrating off Jira with one command: radial import --from jira
A practical guide to migrating from Jira to Radial. Export your project, dry-run radial import --from jira, then bring issues, projects, labels, comments, and history across in a single run.
What is an MCP server? And how to point one at your issue tracker
An MCP server is the thing an AI agent connects to so it can use tools and data outside its own context. Here is what that means, how it differs from a plain API, and how to point one at your issue tracker so your agent files and closes work itself.
Notion is a doc with an AI button. A tracker should stay a tracker.
An honest guide to Notion alternatives in 2026: where Notion fits, the real options for notes and wikis, and why an engineering team tracking issues in Notion should move to a dedicated tracker, one flat locked price, no AI meter.
The best Jira alternative depends on what made you leave
An honest guide to Jira alternatives in 2026: Linear, Plane, GitHub Issues, ClickUp, and where each fits. Then the lane none of them sit in, a fast tracker with one flat locked price and no AI meter.
The meter you didn't sign up for
Usage-based pricing was sold as fairness: pay for what you use. Then the AI era turned every tool into a meter you didn't ask for. Here is what changed, and the case for one flat number that never moves.
The Linear CLI Linear never shipped: driving your tracker from the terminal
Linear has no first-party CLI, so the community built a dozen, most of them aimed at AI agents. Here is the honest roundup, and the case for a tracker that ships the command line as a first-class surface.
Linear vs Jira, and the third option neither names
An honest Linear vs Jira comparison: Linear wins on speed and developer experience, Jira wins on enterprise breadth and customization. Then the question both leave open: why does the price keep moving, and where is the meter?
Give Claude Code access to your issue tracker in 5 minutes (MCP)
Wire Claude Code to your issue tracker over MCP so your agent files, searches, and closes issues itself. A 5-minute setup, the OAuth flow, and why the durable record belongs where the agent already writes.
Why is Jira so slow? A field guide to the 200-requests-per-ticket problem
Jira feels slow because of how it is built: a page that fans out into dozens of round trips before it shows you a ticket. Here is what is actually happening, and what a fast tracker does instead.
Radial vs Linear: same speed, no meter, one locked price
Linear is the best tracker of the AI era, and its agents already ride free. Radial is the one that refused the meter: one flat locked price, a first-class CLI, and a binding pledge against AI you didn't ask for.
Boring on purpose: the tool that disappears is the one that survives 2026
AI fatigue is now a named condition. The counter-trend is calm, durable software that gets out of the way. Here is why a boring issue tracker is the one still standing next year.
Issue tracking isn't dead. Your tracker just got bored of you.
The most-loved issue tracker declared the category over and pivoted to agent management. The category is fine. Someone still has to keep building the fast, boring, durable thing.