How to Use Claude AI: What I Would Do in the First Hour
Sign-up, three settings, the first real job to give it, and what to ignore, with the exact taps on a phone and clicks on a computer.
To start using Claude, go to claude.ai on a computer or install the Claude by Anthropic app on your phone, sign in with Google or your email address, and type in the code it texts to your mobile. The free plan is enough to begin. Before you ask it anything, spend ten minutes in Settings: bigger text, a few lines about who you are, and one privacy switch. Then give it a letter you have read twice and still do not quite understand.
Claude’s screens changed in September and most guides still show the old ones, so everything below uses the words you will actually see.
For my first month with Claude, nearly every chat I started opened with the same sentence. Keep it short, I am reading this on my phone.
When I scrolled back through my history later, 38 of my first 50 chats began that way. The other twelve were the ones where I forgot, got four paragraphs and a bulleted list back, sighed, and typed it anyway.
There is a box in Settings for exactly that sentence. Whatever you write in it applies to every chat from then on. I found it on day thirty-one, looking for something else.
The first hour with Claude is not for asking it questions. It is for telling it who is asking.
Before we get into it: the 10 Claude prompts I use most are free. Join my newsletter at joshdubowsky.beehiiv.com and they arrive the moment you sign up, along with one tested workflow like this each week.
Is Claude free?
Yes, with no end date. The free plan will not run out on you after a fortnight.
The free plan covers chat on the website, the computer app and the phone app, plus web search, memory, voice mode and up to five Projects. What it rations is how much you can do in one go. Every plan has a limit that resets every five hours, the free plan’s is the smallest, and Anthropic says there is no fixed message count. It depends on how long your chats are and what is in them.
Pro is $20 a month in the US, or $17 a month if you pay $200 for the year, with local prices elsewhere. I would not pay on day one. Use the free plan for a week and notice whether you ever hit the limit. If you do, I went through whether Pro is worth $20 in a separate piece.
Step 1: How do I sign up for Claude on a phone or a computer?
Have your mobile beside you either way. You will need it in about two minutes.
On a computer, go to claude.ai. You get two choices: Continue with Google, or type in your email address and choose Continue with email. The email route sends you a message titled “Secure link to log in to Claude.ai”, and you click the link in it. There is no password, and Anthropic says you cannot set one.
One thing caught me out. If you ask for the link on your computer and open the email on your phone, the link shows you a code, and you type that code into the computer. Anthropic’s help page describes exactly this, and I still assumed it was broken.
On a phone, open the App Store or Google Play and search for Claude by Anthropic. Check that the maker is Anthropic. If an iPhone also offers Claude for Intune, leave it, because that one is for large organisations. Tap Get or Install. It needs iOS 18 or later on an iPhone, or Android 8 or later. Open it and you get the same two choices.
On either, it then asks for a mobile number and texts you a six-digit code. Type it in and press Verify code. This is the step that stops some people. Anthropic will not accept landlines, Google Voice, internet phone numbers, numbers created in apps, or anything else that cannot receive a text, and it says there is no way to skip the step. Only three accounts can be verified with one number, and you cannot change the number afterwards, so use the mobile you expect to keep. I think the landline rule is hard on anyone who does not use a mobile, and I could not find a way round it.
Once you are in on one device, sign in on the other with the same email address. Your chats are already there.
If you have a computer, I would do the next three steps on it. Settings are easier to find on a bigger screen, and for one of them Anthropic only describes the computer route.
Step 2: How do I make the text bigger in Claude?
One warning. On 16 September Anthropic merged two parts of Claude into one and began moving paying accounts to the new version, with free accounts to “follow soon”. Until then two people can see slightly different screens, so where a setting lives in two places, I give both.
On a computer, click your initials in the lower left corner, then Settings, and find Appearance. It is under General in mine. Four things in there are worth a look:
- Transcript text size. Small, Medium or Large. This is the conversation itself, and Large makes a real difference on a laptop.
- Chat font. Anthropic’s help page lists Default, Match System and Dyslexic Friendly. Try all three. It takes ten seconds.
- Theme. Light, dark, or matching your computer.
- Motion. Set it to Reduced if things sliding around the screen annoy you. They annoy me.
On a phone, this is where I got cross. That same help page does not mention text size at all, and the app’s App Store page says the developer “has not yet indicated which accessibility features this app supports.” So change your phone’s own setting instead. On an iPhone it is Settings, then Display & Brightness, then Text Size, which Apple says works in apps built to follow it. On Android, open Settings and search for Font size. Then open a chat in Claude and look. If the words did not get bigger, read the long answers on a computer.
One more, on the website. Look under any reply and there is a small row of buttons, one of them Read aloud. It reads the reply out to you. I use it on long answers far more than I expected to, usually while doing something else. (I have only seen it on the website.)
Step 3: How do I tell Claude who I am?
This is the box I took a month to find. Before you fill it in, a question, because I collect the answers:
What is the one thing you wish people would stop doing when they explain something to you?
Keep your answer. It goes in the box.
On a computer, click your initials in the lower left corner, then Settings, and look for Instructions for Claude. In my account it is under Account. Anthropic’s newer help page says General, so check both. Just above it is a smaller box, What should Claude call you? Fill that in too.
Anthropic’s help page for this box only gives the computer route. If you only have a phone, try claude.ai in your phone’s web browser, signed in with the same email. Whatever you write is applied, in Anthropic’s words, “to all of your conversations with Claude”, the ones in the phone app included.
Think of it as briefing someone clever on their first day. They know nothing about you, and they will not ask.
Four lines is plenty. Mine runs to nine, which is too many. Something like this, with your answer to my question on the third line:
I am [ONE LINE ABOUT YOU, e.g. secretary of a walking club].
Keep answers short and in plain English, with no jargon.
Never [YOUR ANSWER, e.g. tell me it is a great question].
Tell me when you are not sure. Do not guess at numbers.Without the last line, it is more likely to hand you a confident figure than to tell you it is guessing.
Claude also builds a memory of its own from your chats, on by default even on the free plan. This box is the part you write yourself, and it works from the very next chat.
Step 4: What should I not tell Claude?
Two switches and one list.
Get Josh Dubowsky’s stories in your inbox
Join Medium for free to get updates from this writer.
The training switch. When Help improve our AI models is on, your chats can be used to train future versions of Claude. On the personal plans, Anthropic’s pricing page lists training as “Opt-out”, which means it is up to you to switch it off. The route is nearly the same on both:
- On a computer: your initials in the lower left corner, then Settings, then Privacy.
- On a phone: tap your initials or your name, then Settings, then Privacy.
Mine is off. I read the whole of what I had agreed to for another piece, and this switch was the part I changed the same day.
The ghost. When you start a new chat outside a project, there is a small ghost icon in the upper right corner. Tap or click it and the chat goes incognito, with a black border and the words “Incognito chat”. It stays out of your history and your memory, and Anthropic does not use it for training even if the switch above is on. Anthropic still keeps it for 30 days, and once you close it you cannot open it again.
The list. Anthropic’s own privacy page asks you to be careful with card and bank account numbers, health records, and “Passwords or private login credentials”. I would add your passport number, and anything with a PIN.
If you want 200 more Claude prompts for jobs like this one, my prompt library is $7 at joshdubowsky.gumroad.com.
Step 5: What should I ask Claude first?
Something real.
The worst first question is a quiz question. What is the capital of Peru, how tall is the Eiffel Tower. It answers, you nod, and you have learnt nothing about whether it is any use to you. A test question gets you a test answer.
Anthropic’s own advice is to “speak to Claude like you would a coworker or friend”, and its newer help page says to describe what you need the way you would brief a colleague: the result you want, the form you want it in, and what it will need from you.
So the first job I would give it is the letter you have read twice and still do not quite understand. Everyone has one somewhere. A renewal, a change of terms, two pages that seem to be saying one thing.
On a phone, start a new chat, tap the + next to the message box, and take a photo of the letter. I fold the top over first, because that is usually where the account number is.
On a computer, click the + in the lower left of the message box, choose Add files or photos, and pick the scan or the PDF. You can also drag the file onto the window, or paste the text straight in.
Then send this:
Here is a letter I have received. In plain English, tell me
what it is asking me to do, by when, and what happens if I
do nothing. Quote the sentence each answer comes from.The quoted sentences are there so you can check each answer in ten seconds instead of taking its word for it.
Whatever it gives you, ask this next:
What did you leave out, and which part of your answer are
you least sure about?It nearly always has an answer. That is the habit I would build in the first week, before trusting it with anything important.
No letter to hand? The email you have been putting off works almost as well. Paste it in and ask for two short replies, one friendly and one firm. Seeing both side by side makes it obvious which bits you actually want.
What can I ignore in Claude for now?
Most of the screen, for the first week. Some of it is good. None of it is needed for a letter.
The model menu. You will notice a menu of models called Haiku, Sonnet, Opus and Fable, which sounds like the contents page of a poetry anthology. They are different versions of Claude. On a computer the menu sits below the message box, and on a phone at the top of the screen. The free plan gets Sonnet and Haiku, and the other two are for paying accounts. Leave it on whichever one it picked, and leave Effort, which sets how hard it works on each answer, alone if you see it there. Nothing that goes wrong in your first week will be fixed by switching.
Most of the + menu. On my account it lists eight things, and this week you need one of them, Add files or photos. It is the busiest thing on the screen and the least use to a beginner.
Skills and Connectors. Depending on your version they sit under Customize in the left-hand menu or in Settings. Skills are add-on instructions for particular jobs, and Connectors let Claude read your Gmail or your calendar. I would give it a month, until you trust its answers, and read what happened when I gave it my inbox before connecting anything.
Projects. These are folders that keep instructions and files together for one ongoing job. Ignore them until you notice you have explained the same job to Claude twice. Then read the five I use, which covers how to make one.
Code. If you see a Code tab, or anything called Claude Code, that is Anthropic’s tool for building software, and a good reason to pay for Pro if that is your work. You do not need it for anything in this guide.
The computer app. There is one for Mac and Windows. The website does everything in this guide, so it can wait.
What do beginners get wrong with Claude?
Three things, mostly, and I did all of them.
Keeping one chat going forever. Every time you send a message, Claude rereads the whole conversation, so long chats use more of your limit. Anthropic’s own advice is to “Try starting a new conversation”. New job, new chat. I wrote a whole piece on what uses the limit up.
Believing the numbers. Claude is right a lot of the time, and it sounds exactly the same when it is not. When Consumer Reports asked chatbots to help choose an SUV, “Claude ranked unreliable vehicles ahead of more reliable ones”. Its knowledge also has an end date. The current models were trained on data up to June 2026, and anything newer it has to look up. Before acting on an answer with a number, a date or a price in it, ask yourself one thing:
If a stranger told me this, would I check it?
If yes, check it.
Expecting pictures. Anthropic says plainly that Claude does not generate photos or illustrations the way image tools do. It can make charts and diagrams. For a picture of your dog in a hat, you need a different tool.
The objection I cannot beat
Everything in this guide makes Claude more useful by telling a company more about you. Your name, your situation, the letters you get, the replies you have been avoiding.
Switching training off helps. Incognito helps. But it still sits on Anthropic’s computers rather than in your desk drawer, and if you have spent twenty years keeping your details off the internet, this article reads like a list of reasons not to start.
I think you get back more than you give, for most letters and most questions. I cannot prove it, and I do not have a good answer for someone who simply does not want to hand it over.
Skip this if
Skip this if you already use Claude most days. You know all of it, and the steps will feel slow.
Skip it if you do not have a mobile that can receive texts. You will not get past sign-up, and I am sorry about that.
Skip it if your Claude comes through work. Your employer’s settings decide most of what is in here.
The thing I would say to someone starting tomorrow
Do the boring ten minutes first. Bigger text, four lines about yourself, one privacy switch. Then give it a real letter instead of a quiz question, and ask it what it left out.
So, what would you hand it first? Tell me in the comments. The most boring answer wins.
P.S. If you would rather have the whole setup than a folder of prompts, The Complete Claude System is at joshdubowsky.com. Memory configured so it remembers you, four projects so nothing gets lost, and five ready-made setups you paste in once. Thirty days, money back, no forms.
