Loading article...
Insights, tutorials, and stories about AI-powered coding and product development
96 articles found
The Claude Code advisor lets Sonnet 5.5 do the routine work while Opus 5.5 advises at the hard moments. Setup, pairing rules, cost illustration and limits.
Sonnet 5.5 nears Opus 5.5 at half the token price, but cheaper per token is not cheaper per task. Here is which model, effort level and ultracode setting to use.
Claude Code mods are plugins that can rewrite prompts, block tool calls and replace built-in features. Here is what changed versus hooks, and the trust decision it creates.
Ultracode is two things in Claude Code: a keyword that runs one task as a multi-agent workflow, and a session toggle that does it for every task. Since September 28 it no longer forces xhigh effort, and it now runs on Claude Opus 5.5 by default. Here is what a real run cost and returned, when it earns its tokens, and how to benchmark it without fooling yourself.
Claude vs Claude Code vs Cowork: same model, different reach. Here is a PM's rule for picking the right one, a comparison table, and a week of PM tasks sorted.
I put Notion Custom Agents, Claude Code Routines, and Grok Bot side by side on user-friendliness, customization, price, and integrations, then tested them with a real Monday brief.
Claude code design just moved inside your terminal. Here is why that changes the Figma question, and who should actually switch now.
Notion's Custom Agents can now run on a model your workspace owner picks. Here's what they do well, what they don't, and how they compare to a Claude Code subagent workflow for PMs.
Most decisions inside an AI feature are not generation, they are routing, scoring, and gating. Here is when Jev's System One model beats an LLM call, and when it does not.
Claude Code swapped Sonnet 5 for Fable 5.1 as its default model on September 1, 2026. Here is what actually changes in cost, benchmarks, and daily use, and whether it is worth caring about.
GitHub Copilot's Agent Merge keeps an AI agent working on a pull request after the code is written, handling review feedback, failed checks, and merge conflicts until it is ready to merge.
Claude API pricing runs $1 to $25 per million tokens depending on model. Here is what it actually costs today, and how Claude Code's new /claude-api hillclimb command finds a cheaper configuration without sacrificing accuracy.
GPT-6 Astra is OpenAI's new flagship model, launched September 3, 2026. Here is what it actually changes for coding, and how it rolled out across Codex, GitHub Copilot, and Lovable.
Cursor Projects introduces a coordinator agent that plans, delegates, and remembers across months of work instead of resetting every session. Here is what it actually does.
Gemini 3.8 Flash is now generally available with the same introductory pricing as 3.7 Flash. Here is what it costs, what changed, and how it holds up on long-horizon coding work.
Codex CLI just added an experimental --worktree flag and Mistral Vibe shipped a full worktree plugin. Here is how both compare to Claude Code's existing worktree workflow for running parallel agent sessions.
Grok Bot's demo is free to watch. Running it on a real business is not. Here is the actual token math, the one-project-per-account rule, and which business models can pay for the $200 to $300 a month.
Google took Gemini 3.5 Transcribe generally available with two dedicated speech-to-text models. Here is what changed and how it stacks up for builders adding voice features.
Omarchy just landed $10 million in foundation funding and a 1,700-plugin ecosystem. Here is whether it is worth switching from a standard Arch setup, for builders, not ricers.
Block's Buzz puts AI agents inside your team chat as first-class teammates, not integrations. Here is what actually makes it different from Slack.
Get the latest articles and tutorials delivered to your inbox every week