AI for Mortals  /  @per_simmons

Claude Sonnet 5 Just Dropped —
Is It Actually Worth It?

Script draft Cold open + outline Run: 2026-06-30

Cold-Open Hook  —  spoken ~145 words

All right. Well, Claude Sonnet 5 just dropped this morning, and as expected, the benchmarks look incredible. On SWE-bench it's putting up numbers that would have made Fable 5 look sideways two months ago. AI Twitter is predictably melting down. "Best coding model at this price tier ever." "Sonnet 5 completely obsoletes GPT-5 Turbo." You know the drill.

But here's the thing — Anthropic priced this at literally the same cost as Sonnet 4. And that is either the most confident move they've ever made, or it's them quietly signaling something. So I spent about three hours this afternoon stress testing it on the actual builds I run every day, side by side with what I was already using. And I'll tell you: the results were not what I was expecting — in a good way.

In this video, I'll show you exactly what's actually different, which tasks it completely blows Sonnet 4 out of the water on, and whether for most of you, this is now the model you should have running 24/7.

≈ 148 words

Section-by-Section Outline  —  7 sections

  1. Where Sonnet 5 Actually Sits in the Lineup Quick map of the current tier stack — Haiku, Sonnet, Fable, Opus — so viewers understand what they're buying into and why the same price point as Sonnet 4 is actually a weird flex worth unpacking.
  2. The Benchmarks: What to Actually Care About Walk through the numbers that matter, skip the ones that don't, and flag the one metric where the jump from Sonnet 4 is not hype — plus the one where the benchmark is doing a lot of heavy lifting for Anthropic's PR team.
  3. Everyday Work Test: Side by Side Sonnet 4 vs. Sonnet 5 on the tasks most of the audience actually runs daily — drafting, summarizing, reasoning through a real problem. Where you'll feel the difference. Where you honestly won't.
  4. The Coding Test — Where It Earns It One-shot builds, agentic tasks in Claude Code, a complex refactor. Sonnet 5 is noticeably better on the hard stuff — and there's one honest miss where Sonnet 4 held its own that's worth flagging before you go all in.
  5. The Cost Math: Same Price, But Compared to What? Breaking down Sonnet 5 vs. Sonnet 4 cost-per-output, and then the more interesting question: does this become your default over Fable 5 credits, or is it a stepping stone? Running the actual numbers on a real workflow.
  6. The Verdict: Who Should Actually Switch Today Honest breakdown by use case — power builders, casual Claude users, Claude Code daily drivers, and people still on Sonnet 4 — who flips the switch now and who honestly can wait. No hype, just a real recommendation.
  7. The Thing Nobody Is Talking About Something Anthropic buried in the release notes that almost everyone missed — a quiet change in how Sonnet 5 handles long-context agent loops that actually matters more day-to-day than the benchmark jump. Worth staying for this one.

AI for Mortals — simmonsbench / 2026-06-30