OPUS 5 first results:
WOZCODE vs Claude Code — fixing bugs in an Express + TypeScript API (6 real prompts):
💰 $13 vs $21 (41% cheaper)
🔁 96 vs 227 turns (58% fewer)
⏱️ 31m 11s vs 36m 44s (15% less time)
⭐ quality 8.1 vs 8.1 (of 10)
Full numbers: wozcode.com/benchmark-histor…
Kimi K3 BEATS Claude Code on our benchmarks.
6 real bug-fix & feature tasks on an Express + TypeScript API, blind-judged:
🏆 Quality: 8.1 vs 7.8 — the open-weights model outscores stock Claude Code running Claude Fable 5
💰 Cost: $8.69 vs $31.19 — 72% cheaper
🔁 Turns: 101 vs 174
⏱️ The tradeoff: ~2× slower (89m vs 46m)
Full numbers: wozcode.com/benchmark-histor…
GPT5.6 Sol
WOZCODE (in Codex) vs Codex — fixing bugs in an Express + TypeScript API (6 real prompts):
💰 $4.74 vs $8.50 (44% cheaper)
🔁 18 vs 23 turns (22% fewer)
⏱️ 6m 58s vs 14m 50s (53% less time)
⭐ quality 7.8 vs 7.5 (of 10)
Full numbers: wozcode.com/benchmark-histor…
WOZCODE vs Claude Code — building a todo app with local storage (30 real prompts):
💰 $6.93 vs $10 (31% cheaper)
🔁 78 vs 119 turns (34% fewer)
⏱️ 14m 34s vs 19m 28s (25% less time)
⭐ quality 9.4 vs 9.1 (of 10)
Full numbers: wozcode.com/benchmark-histor…
FABLE 5:
WOZCODE vs Claude Code — screenshot-driven art portfolio builds (10 real prompts):
💰 $13 vs $31 (57% cheaper)
🔁 42 vs 202 turns (79% fewer)
⏱️ 31m 7s vs 45m 14s (31% less time)
⭐ quality 8.1 vs 7.3 (of 10)
Full numbers: wozcode.com/benchmark-histor…
FABLE 5:
WOZCODE vs Claude Code — fixing bugs in an Express + TypeScript API (6 real prompts):
💰 $15 vs $33 (56% cheaper)
🔁 51 vs 240 turns (79% fewer)
⏱️ 27m 59s vs 44m 24s (37% less time)
⭐ quality 8.1 vs 7.7 (of 10)
Full numbers: wozcode.com/benchmark-histor…
WOZCODE vs Claude Code — fixing bugs in an Express + TypeScript API (6 real prompts):
💰 $10 vs $23 (54% cheaper)
🔁 91 vs 199 turns (54% fewer)
⏱️ 35m 26s vs 49m 39s (29% less time)
⭐ quality 7.8 vs 7.6 (of 10)
Full numbers: wozcode.com/benchmark-histor…
Woz retweeted
Can't make the case for WOZCODE any better than this.
Companies need to be able to control costs while also leveraging frontier models
Our Anthropic bill is about to jump from $400K → $1.4M/yr.
Not because usage exploded, but because we're about to cross 150 seats.
Past 150 seats you're forced into Enterprise tier. Seats stop including any usage, every token bills at standard API rates. At our current run rate that's 3.5x overnight.
Unfiltered thoughts on AI spend:
1. We should spend tokens to grow as aggressively as possible. But most people (me included) aren't conscious of what they're spending.
2. Visibility comes first. People see their personal number and they're shocked. I accidentally spent $4,000 in 3 days in Claude Code.
3. For engineering the spend is clearly worth it. Pay for the best model, it saves more than it costs.
4. For a lot of other roles it's questionable. Apps nobody uses, skills someone already built. No ROI.
5. Spend limits are coming. We already require approval for more tokens on our support team.
The era of token-maxxing is coming to an end.
Woz retweeted
Replying to @_catwu
We've built replacements for Claude Code's core tools (Read/Write) that are significantly more efficient, would love to be able to give our customers a deeper and more seamless integration
WOZCODE reads files differently.
Instead of dumping entire file contents into context, AST truncation replaces function bodies with stubs. Types, exports, and signatures stay intact.
Same useful information. 40 to 60% fewer tokens.
And because every token gets re-ingested on every subsequent call, a leaner read compounds across the entire session.
Built in. No configuration.
Woz retweeted
Replying to @yacineMTB
We built a recall tool that does essentially fuzzy finder on all previous Claude sessions and it’s been a total gamechanger
/woz-recall how did we fix that caching bug last week?





