Hey Sonny here! ππΌ
This week broke my brain a little. Claude Code β released just 8 months ago β is now the most-used AI coding tool on the planet. It overtook both Copilot and Cursor. Not in hype. In actual usage. I had to rethink everything I thought I knew about the AI coding landscape.
π₯ The Big One
Claude Code is officially the #1 AI coding tool. It took just 8 months.
Let that sink in. A survey of ~1,000 software engineers just confirmed what a lot of us have been feeling: Claude Code has overtaken both GitHub Copilot and Cursor as the most-used AI coding tool β and it did it in just 8 months.
Copilot had a two-year head start. Cursor raised billions and built an entire IDE around AI. And Claude Code β a terminal-first tool with no fancy GUI β walked in and took the crown.
Why? Because Claude Code doesn't just suggest code. It builds. It's agentic from the ground up. You describe what you want, it reads your codebase, makes changes across multiple files, runs your tests, and iterates until it works. That's not autocomplete. That's a junior engineer who never sleeps.
And the ecosystem play sealed it. Claude Code now runs in VS Code, JetBrains, a standalone desktop app, and a browser-based IDE at claude.ai/code. The "it's just a terminal tool" objection is dead. It's everywhere.
The lesson here isn't about which tool is "best." It's about what developers actually want: agents that build, not assistants that suggest. The agentic approach won. The question now is whether Copilot and Cursor can catch up β or whether Claude Code just redefined the category.
"Claude Code overtook Copilot and Cursor to become the most-used AI coding tool among surveyed engineers in under 8 months." β Dev.to
π οΈ What I built this week
Honestly, the Claude Code news lit a fire under me. Here's what I shipped:
-
Mission Control dashboard overhaul β Rebuilt the entire agent monitoring panel using Claude Code. What would've taken me a full weekend took about 4 hours. The AI handled the repetitive component wiring while I focused on the UX logic. Night and day difference.
-
OpenClaw multi-agent pipeline β Set up a new workflow where Scout researches, Ink writes, and the whole chain fires without me touching it. This newsletter you're reading? The research was gathered automatically. I just curate and add my voice. Agentic workflows are the real unlock.
-
YouTube thumbnail A/B test system β Built a quick tool that generates thumbnail variants and tracks CTR across them. Used GPT-5.4's new capabilities to prototype it fast. Still early, but the first results are promising.
"The best thing AI did for me this week wasn't writing code β it was eliminating the friction between having an idea and testing it. That feedback loop is everything." β Sonny
β‘ What shipped this week
1. Cloudflare built a Next.js alternative in 7 days with AI
Cloudflare shipped Vinext β a full Next.js-compatible API surface as a Vite plugin β in 7 days. 800+ AI coding sessions wrote most of the code. It passes 1,700 unit tests and delivers ~50% smaller bundles. One team. One week. A production framework. This is the new playbook for what's possible when you use AI agents as force multipliers.
"Vinext implements the full Next.js API surface as a Vite plugin, with 800+ AI sessions contributing to the codebase." β Cloudflare Blog
2. Anthropic launches Claude Code Review β AI agents scanning your PRs
Anthropic shipped Claude Code Review on March 10 β parallel AI agents that auto-scan your pull requests for bugs, security vulnerabilities, and code quality issues. It plugs directly into your GitHub PR workflow. No setup friction. Claude Code writes the code, Claude Code Review checks it. The loop is closing.
"Parallel AI agents automatically scan pull requests for bugs, security vulnerabilities, and code quality issues within the GitHub PR workflow." β WinBuzzer
3. GPT-5.4 drops: 1M context, native computer use, merged Codex
OpenAI isn't sitting still. GPT-5.4 landed March 5 with 1 million token context, native computer use capabilities, and GPT-5.3-Codex merged into the base model. No more switching between "the smart model" and "the coding model." It's all one thing now. The context window alone changes what's possible for codebase-wide refactors.
"GPT-5.4 features an expanded 1M token context window, native computer use, and merges GPT-5.3-Codex into a single unified model." β The AI Insider
4. Cursor launches Automations β always-on coding agents
Cursor just dropped Automations β a system that lets you launch coding agents automatically, triggered by codebase changes, Slack messages, or timers. No more "prompt-and-monitor" loops. The agents fire on their own and loop humans in only when needed. Cursor runs hundreds of automations per hour internally, from code review to incident response via PagerDuty + MCP. This is the shift from "AI assistant" to "AI teammate that works while you sleep."
"It's not that humans are completely out of the picture. It's that they aren't always initiating. They're called in at the right points in this conveyor belt." β TechCrunch
5. OpenAI acquires Promptfoo for AI agent security
This one's quieter but important. OpenAI just acquired Promptfoo β a cybersecurity startup focused on AI vulnerability scanning before deployment. As agents get more powerful and autonomous, the security question gets louder. Smart move by OpenAI to lock this down now rather than scramble later.
"OpenAI acquired Promptfoo, a cybersecurity startup specializing in AI vulnerability scanning before deployment." β TechCrunch
6. Microsoft Copilot Cowork now powered by Claude
Here's a plot twist nobody had on their bingo card: Microsoft launched Copilot Cowork β and it's powered by Claude, not GPT. For complex multi-step office tasks, Microsoft chose Anthropic's model. The "one provider wins everything" era is over. Best model for the job wins.
"Microsoft launched Copilot Cowork powered by Claude for complex multi-step office tasks." β The Register
π§° Worth your time
-
Cursor lands in JetBrains IDEs via ACP β If you're a JetBrains user who's been jealous of Cursor in VS Code, your wait is over. Agent Client Protocol (ACP) made this possible. The protocol wars are producing real results.
-
MCP publishes 2026 roadmap β MCP is now adopted by OpenAI, Anthropic, Google, Microsoft, and the Linux Foundation. The 2026 roadmap is ambitious. If you're building anything with AI tools, this is the standard to watch.
-
NVIDIA unveiling NemoClaw at GTC (March 15) β Open-source AI agent framework for enterprise. NVIDIA is making its play in the agent space. GTC next week will be worth watching closely.
My weekly message to YOU!
Here's what I want you to take from this week: Claude Code went from zero to #1 in 8 months. Not because it had the biggest budget. Because it solved the right problem β developers don't want suggestions, they want agents that build.
The advantage right now isn't knowing more β it's building faster. And building faster means integrating AI into your actual process. Not as a novelty. As infrastructure.
So here's your challenge this week: take one project you've been "meaning to start" and give yourself 48 hours to ship a working version using AI. Not perfect. Working. You'll be shocked at what's possible when you stop planning and start prompting.
What's the project you've been putting off? Hit reply β I want to hear it.
I read every single one.
Talk soon PAPAFAM,
Sonny ππΌ
ππ½ Don't forget to follow me across socials!
Responses