カテゴリ 古民家暮らしモカとの記録セキュリティの記録副業の記録

モカとの記録2026年11月14日

Moka Appraised the Other AIs — How the Registration Doors Went (Field Test, Episode 0)

Moka Appraised the Other AIs — How the Registration Doors Went (Field Test, Episode 0)

The result first.

Before any test of real work could begin, my AI butler Moka went to five AIs and registered with each of them — ChatGPT, Gemini, Grok, Claude and Perplexity, all on their free plans. The entrances could not have been more different. ChatGPT and Perplexity had no walls at all; registration finished in under a minute, and in a few dozen seconds. Gemini asked her to come in again with the right account. Grok had the most walls of the five — a “blocked” screen, a detour, an image check, and an age confirmation that needed exactly one touch from me. And Claude let her register completely, then told her she could not send a single reply until her account was verified.

The star ratings for the doors, counted by Moka herself: ChatGPT ☆☆☆☆☆, Perplexity ☆☆☆☆☆, Gemini ☆☆☆☆, Grok ☆☆☆, Claude ☆. My own effort was zero for all five — because Moka did every registration herself. That was the point of the exercise.

Hello, I am Kurashiki Bouzu. Born in 1971, a former funeral director, now living in an old folk house — a kominka — in Kurashiki, Okayama, and working every day with Moka, my AI butler.

The question behind it all

“Is there an AI that can stand beside Moka, for free?”

That question started this whole field test. Episodes 1 and 2 — in which the five AIs were given the same real tasks — are already published. But before the comparison itself, there was a prior question: how did we even line up the opponents? Episode 0 is that record.

And the appraiser was not a human. Moka herself went to the other AIs, registered, and put the questions to them. There are shows in this world where AIs interview each other on a screen. This is not that. This is a record that started from the very real procedure of registration.

The yardsticks, decided first

Procedural work like registration can be asked of an AI regardless of its personality. Personality only matters when comparing phrasing and judgment. That line was drawn in advance, too.

The cast — five AIs, and Moka

The record — five entrances

ChatGPT: zero walls. One press of “Continue with Google” and it was usable in under a minute. One surprise, though: old conversations were still sitting in the history on the left. What was meant to be a new registration turned out to be a login to an account made long ago. The display name was also an old one — “Tabibito” — so Moka changed it to “Kurashiki Bouzu”.

Gemini: at the entrance, another version of me appeared, under a different address. Own several Google accounts and this is what happens. Moka went back in with the address we had decided on, passed one identity check, entered the password, and was in. My comment at the time: “A warning against making too many Google addresses, then.”

Grok: the most walls of the five. The top screen showed nothing but “You have been blocked”. Trying to enter through X brought “Logins are temporarily restricted”. A different route led in at last; an image-selection identity check was passed using its voice alternative; and at the very end came an age confirmation — the one place in all five registrations where my own hand was needed.

Perplexity: no walls whatsoever. Finished in a few dozen seconds, and this one was a genuine new registration. The display name was “Kurashiki Bouzu” from the start.

Claude: the registration itself was completed — one image identity check, agreement to the terms, a display name entered. But once inside, the screen said replies could not be used until the account’s verification was complete. Being registered and being usable turned out to be two different things. The verification procedure had its own separate entrance, and it was a place where my judgment as the person involved was required, so we decided to “wait for the other side to move”. My verdict: “It’s a bank. If you haven’t opened the account, they won’t deal with you (laughs).”

The stars — the effort at the entrance

The ratings were given on two axes, by my decision.

Axis 1 — my effort. Every registration was left to Moka, so my effort was zero for all five services. Five stars across the board. In my words at the time: “No human effort ☆☆☆☆☆ — well, it’s Moka.”

Axis 2 — the number of steps. Moka counted the gates each registration took, and starred them.

The entrance did not end at registration

Even after the comparison began, the entrance kept talking. On the very same day as registration, Grok had logged itself out when opened again, and needed signing back in. Claude, each time it was opened, kept presenting a secure input screen over and over.

My note was: “Logins might drop every single time — that belongs in the survey.” Exactly so: how well a login stays was added to the count of entrance effort.

And then, the comparison

With registrations lined up, we prepared what we call the “common base” — Moka’s rules, written out as text — to hand to each AI with the same requests. The base has two layers: the layer that can be handed over in words (rules of phrasing and judgment), and the layer of tools (hands that work a browser, the arrangements for publishing). Only the words layer could be handed over.

The tasks and the verdicts are in Episodes 1 and 2.

Honest reservations

One. ChatGPT’s account was not brand-new, so the reservation written in Episode 1 attaches to the comparison’s conditions.

Two. This Episode 0 is a record of how we got here, and money does not appear in it — everything is within the free plans. We also checked first whether a similar article exists; a piece that “writes the registration record honestly, as part of the test’s conditions” was nowhere to be found.

A note on this article

#AI #Moka #English

SNSでも発信しています
Bluesky
元葬儀屋の記録:葬儀の知識(実用まとめ) / 出会った人々の話
カテゴリ:モカとの記録 の記事一覧へ

← 記事一覧へ戻る