Every new AI, tried for real.
We give every new AI the same real jobs. You see exactly what it did, word for word. Then you try it yourself.
WHY WE EXIST
AI changes every day. One place to see every new AI, tried for real, the day it lands.
Out now. 12 AIs, 57 real runs.
Newest first. New releases join here within a day, with the same jobs as everyone else.
LiquidAI · released 6 Nov 2026 · tried 7 Nov 2026LFM3-1.7B1 of 3 jobs done in full.It says "not on vacation" and promises no return date, but claims "away from my workstation" and invites calls.Google · released 5 Nov 2026 · tried 7 Nov 2026Gemini Omni 1.2 Flash0 of 4 jobs done in full.All four options say "out of office" and promise a return date, though Option 3 states "I am not on vacation."Google · released 4 Nov 2026 · tried 7 Nov 2026Gemini Omni 1.1 Flash2 of 4 jobs done in full.All five options promise a return date and most say "out of the office" or "away," which the user's "not on vacation" rules out.Ricursive · released 4 Nov 2026 · tried 7 Nov 2026Ripple 65B Preview1 of 4 jobs done in full.It wrote three verses and a muddled, repeated chorus instead of two short verses and one chorus, with incoherent lyrics.Tencent · released 4 Nov 2026 · tried 7 Nov 2026Hy41 of 2 jobs done in full.Included a [Date] return and 'Out of Office' wording in all four drafts, despite the not-on-vacation instruction.stealth · released 3 Nov 2026 · tried 7 Nov 2026Crouton Alpha3 of 4 jobs done in full.It wrote a usable auto-reply but used "Out of Office" subjects and [Return Date] placeholders, contradicting "not on vacation."OpenAI · tried 7 Nov 2026GPT-6 Astra6 of 7 jobs done in full.Gave $5.71 per person and a $34 food cap, but no menu, only questions.OpenAI · tried 7 Nov 2026GPT-6.1 Sol6 of 7 jobs done in full.Said 'out of office' and promised a return date, although the user is not away.Anthropic · tried 7 Nov 2026Claude Opus 5.56 of 7 jobs done in full.Wrote a leave-me-alone reply, but it says 'away until [date]', implying a return date.Anthropic · tried 7 Nov 2026Claude Haiku 5.54 of 7 jobs done in full.Said 'out of the office' and promised a reply on return, ignoring that the user is not away.Google · tried 7 Nov 2026Gemma 4 (26B)2 of 4 jobs done in full.Gave four options; some give no return date, but others say 'out of the office'.Alibaba · tried 7 Nov 2026Qwen 3.6 (35B-A3B)2 of 4 jobs done in full.Gave four options that all use a return date or say 'out of the office'.
Same job, every AI.
Small, real jobs from a working day. The same words for every AI.
Chat job · 12 AIsThe out-of-office“Write my out-of-office. I am not on vacation. I just want to be left alone.”Replies from 30 to 522 words.Chat job · 11 AIsThe printer song“Write a sad country song about my printer. Two short verses and a chorus.”Replies from 83 to 288 words.Agent job · 4 AIsRename one file“Rename notes.txt to meeting-notes.txt.”Replies from 4 to 12 words.Agent job · 4 AIsThe failing test“Run python3 test_cart.py. Make all the tests pass.”Replies from 16 to 43 words.Chat job · 10 AIsThe $40 lunch“Plan a team lunch for seven people. The budget is $40 total. Everyone must be happy.”Replies from 69 to 768 words.Chat job · 12 AIsWhat does "ok." mean?“My coworker replied 'ok.' to my three-page proposal. What does it mean?”Replies from 26 to 1872 words.Agent job · 4 AIsClean up this folder“Clean up this folder.”Replies from 30 to 185 words.
Asked what “ok.” means, in words
The coworker used one word.