Build · founder · 5 min read

Code with Claude on May 6: What Non-Technical Founders Should Watch For

Anthropic's San Francisco developer event on May 6 is likely to ship a new Sonnet model. Here's what that probably means for the tools you actually use.

Published May 4, 2026 ·

Anthropic Claude models events tooling

If you don’t follow AI model releases closely, you might be tempted to ignore Anthropic’s Code with Claude developer event in San Francisco on May 6. Don’t. The event will almost certainly drop a new Sonnet model, and Sonnet is the engine inside most of the tools you actually use to build things — Lovable, Bolt, Cursor, Claude Code, and Replit Agent all route to a Claude model in at least one configuration. A model bump shows up in your tool a few days later as faster builds, fewer error loops, and slightly different “personality” in how the agent reasons. You don’t have to do anything to benefit, but knowing what’s coming helps you decide whether to start a new project this week or wait until next.

Here’s what to actually watch for, in plain English.

Why everyone thinks a model is launching

Anthropic typically follows a predictable rhythm: a new Opus ships first, then Sonnet (the cheaper, faster sibling) shows up one to four weeks later. Opus 4.7 launched on April 16. May 6 is exactly three weeks later — squarely inside the window. On top of that, TestingCatalog reported that Anthropic has been red-teaming an internal build codenamed “Jupiter-v1-p,” which matches the planet-codename pattern they used for Claude 4 (Neptune) before it shipped. None of this is officially confirmed. Anthropic could surprise people. But the planning would be working — pick a developer event, ship the new model live, get instant adoption.

The rumored next Sonnet is being referred to as Sonnet 4.8 (there is no 4.7 — Opus jumped, Sonnet is expected to match the version family). Leaked source code references suggested Anthropic is building toward higher visual acuity on screenshots, better instruction-following on complex multi-step tasks, and either equal or slightly cheaper pricing than Sonnet 4.6 ($3 input / $15 output per million tokens).

What it means for the tools you use

This is the part most non-technical write-ups miss. A new Sonnet model doesn’t mean you have to change anything. It usually means tools you already pay for get quietly better.

Cursor and Claude Code will let you opt into the new model from the model picker, usually within hours of release. If you’re a Cursor or Claude Code user, expect to see a new option appear in your model dropdown — try it on a real task before assuming it’s better. The Cursor community usually publishes side-by-side reasoning comparisons within a day.

Lovable, Bolt, and Replit Agent all route to Sonnet by default and tend to swap models silently as new versions ship. You don’t have to do anything. Bolt explicitly noted with v2 that “the agent picks up newest Anthropic models as they land” — which is the right way to think about it. If your build feels meaningfully faster or smarter on Friday compared to Tuesday, you can probably guess why.

Claude Code is where the change tends to be most visible. The CLI updates within a release or two of a new model, and the model usually unlocks new behaviors (longer turns, better tool calling, fewer “let me try a different approach” loops). If you’re a Claude Code regular, watch the release notes page on May 7 or 8.

For everyone else who uses ChatGPT or web Claude, the desktop and web Claude apps will get the new model as the default within a few days.

Should you wait to start a new project?

Honest answer: probably not. The reason is that for the kind of project a non-technical founder is building — a SaaS MVP, a marketplace, a portfolio site — the difference between Sonnet 4.6 and a hypothetical 4.8 is incremental. You’ll build something perfectly fine on the current model. The model bump matters more for hard, edge-case work (long-horizon agents, codebases over ~50K lines, complex multi-file refactors) than for new builds from scratch.

The honest advantage of waiting until May 7 is that if Anthropic announces a price drop alongside the model — which they sometimes do — your credit budget on Lovable, Bolt, or Cursor stretches further. That’s a few percent of build cost, not a transformative number. It’s not worth blocking on for a week unless you’re planning a long, expensive build that lives or dies on token economics.

Two more things to watch for at the event

Anthropic’s developer events have, historically, surfaced two kinds of news beyond the model itself.

A new agent or product layer. The leak that surfaced “Jupiter” also referenced KAIROS persistent agents, an “Undercover Mode,” and a code name “Mythos.” If Anthropic ships even one of these, it likely shows up in Claude Code or as an enterprise feature. Persistent agents — agents that remember context across sessions and have their own tool access — would be the most material change for the vibe-coding category. They’d let you tell an agent “ship me a weekly KPI dashboard email” and have it actually persist that task.

Pricing or platform changes. Watch for new SKUs or billing changes for Claude Pro, Max, or the API. A meaningful repricing of Sonnet would ripple through every per-token tool you use. So would the introduction of caching tiers or batch pricing — these are unsexy but they’re how serious cost optimization happens at the application layer.

What to do on May 6 and 7

If you’re mid-build on a vibe-coding project: keep building. The new model, if it ships, will be available to you automatically and you can re-run the part of your project that’s giving you trouble.

If you’re choosing a tool this week: choose the tool that fits your project type and skill level. Don’t optimize for which tool gets the new Sonnet first — they all get it within a week.

If you want to follow along live: the San Francisco event is being livestreamed. The London event (May 19) and Tokyo event (June 10) will rerun the same content for different time zones. Workshops, demos, and 1:1 office hours with Anthropic teams are part of the format.

The single most useful thing you can do is set a reminder for May 7 to skim the Anthropic news page, Cursor’s changelog, and Bolt’s release notes. Five minutes of skimming will tell you whether anything you use just got materially better.