{"id":559,"date":"2026-10-26T16:56:28","date_gmt":"2026-10-26T15:56:28","guid":{"rendered":"https:\/\/erikagherardi.it\/lab\/english-companion-the-english-tutor-that-gets-you-talking\/"},"modified":"2026-10-26T17:35:26","modified_gmt":"2026-10-26T16:35:26","slug":"english-companion","status":"publish","type":"lab_project","link":"https:\/\/erikagherardi.it\/en\/lab\/english-companion\/","title":{"rendered":"English Companion, the English tutor that gets you talking"},"content":{"rendered":"

The need<\/h2>\n

My partner understands English well but speaks it little. After months of Duolingo without much to show for it, I decided to build her English Companion, a GPT that is in effect a personal tutor, helping build confidence with everyday conversations and levels that get gradually harder.<\/p>\n

How I built it and how it works<\/h2>\n

I started with ChatGPT, first of all asking it how a GPT works and how to create one, and then we began shaping the structure starting from the needs. The result was decent, but the process was frustrating: ChatGPT kept adding new ideas instead of building, and creating the files turned out to be more cumbersome than expected.<\/p>\n

So I gave everything to Claude, asking it to evaluate the structure against the purpose. The result was a general overhaul of the system and finally 15 core knowledge files that hold the whole method, from the learning path to the outline of each level, while the instructions tell ChatGPT how the tutor behaves and which file to use, and when.<\/p>\n

That is because a custom GPT has both a space for instructions and a space where you can attach files that let you extend how it works. The logic followed was to put in the instructions the directions for using those 15 files, which are the beating heart of the system.<\/p>\n

\n
\"The\"The<\/div>
The 15 knowledge files that make up the method, and the five conversation starters.<\/figcaption><\/figure>\n

But what does it actually do? ChatGPT behaves like a coach who asks you simple questions and keeps the conversation going. First it lets you answer, then it explains what is right, what is wrong and what could be better. Italian words are welcome when English is not enough. It is patient: it lets you finish your sentence and ends every correction on something that went well.<\/p>\n

\n\"The<\/a>
The start of a first mission: a warm welcome, a reminder that the goal is to communicate and not to be perfect, and a warm-up question with an example to follow.<\/figcaption><\/figure>\n

Underneath there is a system of levels: increasingly complex topics, but always drawn from everyday life and work. Five starter phrases open the most common situations, from the mission of the day to preparing for a meeting, and, if you do not change window, the system also remembers where you are. In a new one you just paste the summary it hands you at the end of the session.<\/p>\n

\n\"A<\/a>
The passport summary at the end of a session: where you got to, what went well and what to review. It can be pasted into the next chat to pick up where you left off.<\/figcaption><\/figure>\n

Where it stands<\/h2>\n

English Companion works and is open to anyone with the link, but it will soon change format: OpenAI is asking GPT creators to turn them into plugins, and English Companion will have to do the same.<\/p>\n

Has it been useful? To be honest, my partner has used it no more than a couple of times, while I still use it today to practise.<\/p>\n","protected":false},"excerpt":{"rendered":"

A personal English tutor built as a custom GPT and focused on conversation. You grow through a system of levels, from the basics to topics from everyday life and work.<\/p>\n","protected":false},"featured_media":0,"template":"","project_status":[19],"project_type":[27],"class_list":["post-559","lab_project","type-lab_project","status-publish","hentry","project_status-built","project_type-ai"],"_links":{"self":[{"href":"https:\/\/erikagherardi.it\/en\/wp-json\/wp\/v2\/lab_project\/559","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/erikagherardi.it\/en\/wp-json\/wp\/v2\/lab_project"}],"about":[{"href":"https:\/\/erikagherardi.it\/en\/wp-json\/wp\/v2\/types\/lab_project"}],"version-history":[{"count":1,"href":"https:\/\/erikagherardi.it\/en\/wp-json\/wp\/v2\/lab_project\/559\/revisions"}],"predecessor-version":[{"id":573,"href":"https:\/\/erikagherardi.it\/en\/wp-json\/wp\/v2\/lab_project\/559\/revisions\/573"}],"wp:attachment":[{"href":"https:\/\/erikagherardi.it\/en\/wp-json\/wp\/v2\/media?parent=559"}],"wp:term":[{"taxonomy":"project_status","embeddable":true,"href":"https:\/\/erikagherardi.it\/en\/wp-json\/wp\/v2\/project_status?post=559"},{"taxonomy":"project_type","embeddable":true,"href":"https:\/\/erikagherardi.it\/en\/wp-json\/wp\/v2\/project_type?post=559"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}