Almost certainly you read the first edition of the newsletter and gave Claude proper, solid prompts all this week. Congratulations! You actually don't need to read anymore, you're officially on your way to mastery.
Me? I don't like settling, so let's build on what we covered and get even better results this week.
Picking Up Where We Left Off
In the first edition we covered what's needed up front for Claude to produce better results. We framed is as a template (Context, Example, Instructions, Constraints, Output Format) and the idea that answering questions before they're asked will lead to better outputs right away.
That's still true, and if you've been practicing it this week, your first responses from Claude are probably noticeably better.
However, that's just the beginning (pun intended). What I didn't tell you last week:
A well-prompted first response is still a first draft.
This week we're going to build on top of the first edition and talk about the difference between a first draft and a finished product.
What Anthropic Found When They Watched
Earlier this year, Anthropic (Owners of Claude) studied nearly 10,000 conversations on Claude to figure out what separates people who get great results from people who don't.
Their results showed that the biggest factor was what happened after the first response. People who pushed back, refined, and iterated on Claude's output showed roughly double the number of AI fluency behaviours compared to people who accepted the first response and moved on. Generally they:
- Caught more gaps
- Questioned more assumptions
- Corrected more tone & behaviour
- Steered more outcomes
In doing so, they ended up with output that was meaningfully better, not just slightly different.
That hit me. When I started using AI tools I was absolutely one of the people who would get a decent first response and move on. Many of us do this for many reasons, whether it's from not really knowing better or simply too busy and need to move on to the next thing.
The problem is that it compounds. One decent response used for an email is fine, but a hundred of them start to look like someone else wrote them. Probably because someone else did.
The Skill: Giving Quality Feedback
The skill to learn is giving quality feedback. If we're just saying "try again" and hoping for something better, well, that's like telling a teammate "this isn't right" without explaining what's off. Don't be that person.
Quality feedback is being specific about what needs to change. Outlining what may be off, and guiding with some context to correct it. Here are some examples of feedback I've personally given in the past month:
The tone is too formal. I need this to sound like I'm talking to a client I've known for years, not writing a press release
The second paragraph is stronger than the first. Rewrite the opening to match that energy
You're leading with my credentials. Lead with the problem I solve instead
This is 300 words. I need it under 120. Be more concise and direct with the opening statement.
The current approach is safe but forgettable. Give me two rewrites to A/B test, one that is a bit more edgy and one that is takes a completely different angle.
What these have in common is that they don't just say "it's wrong." They point at the specific thing that's off and give Claude a direction to move in. Tone, structure, length, angle, these are examples of a few levers that can help reshape the response.
When It Gets Worse Instead of Better
I need to be honest about something, this doesn't always work cleanly.
There's a real failure mode where you iterate too many times in one conversation and the output starts degrading instead of improving. I know we've all been there, you give feedback, Claude adjusts, you give more feedback, Claude adjusts again, and somewhere around the sixth or seventh round, the result is worse than what you had three rounds ago.
There is a mental model to help with this. Think of it like a whiteboard that never gets erased. Every note, every crossed-out idea, every piece of conflicting feedback stays visible. Claude has to write its next response while staring at all of it at once. Eventually the signal gets lost in the noise.
The practical sweet spot I've landed on is roughly three to five rounds of meaningful feedback. That's not a hard rule, but it's been a useful threshold for me. If I'm approaching three to five rounds and it's getting muddier, that's my cue to stop iterating in that thread and realizing; I will not find Pepe Silvia here.
Start a new conversation, take the best version you have so far and paste it into the fresh chat and give Claude a clean brief.
Here's a draft I've been working on. Here's what's working: [X]. Here's what needs to change: [Y]. Give me a revised version.
That reset almost always produces a better result than round seven in a tired thread.
Does It Really Make That Much of a Difference?
Honestly? I wondered the same thing early on. The Anthropic research I mentioned answers it pretty clearly. The people who iterated caught gaps and questioned assumptions at several times the rate of people who didn't. The improvements weren't marginal, they were significant.
But the part that matters more to me isn't any single output being better. It's what happens over time. Every time you push back and refine, you're getting sharper at knowing what to ask for in the first place. You start to internalize what "good" looks like for the types of output you create regularly. Your first prompts get better because you've seen the gap between a first draft and a finished one enough times to close it proactively.
Let's not forget that Claude is learning too. Over time, it picks up on your patterns, your preferences, how you like things structured and worded. The collaboration with the tool just becomes more and more efficient. So, think of it like iteration today is front-loading for tomorrow.
If that sounds like it connects to last week's edition, it does. Front-loading context and iterating on output are two halves of the same skill. One is about input. The other is about refinement. Together, they're how you go from "using Claude" to working with it.
Crushing Busywork
Let's crush some busywork now with this new skill. Here are three things you may have done this week that took longer than they needed to.
The end-of-day client recap
You spent fifteen minutes writing a summary email to a client after a meeting or a session or god forbid, your company uses Teams and you were given a Co-Pilot summary... This week, give Claude your rough notes and ask for a recap email. When it comes back, find opportunities to provide a couple of quality feedback, try something like:
Too formal. This client and I have a good relationship, make it conversational but keep it professional. Also, move the next steps to the top, that's what they care about.
Two rounds of feedback and you're done in three minutes instead of fifteen.
The internal memo nobody wants to write
A policy changed, a process shifted, whatever the reason, something needs to be communicated to the team. You know what you need to say but drafting it feels like pulling teeth. Give Claude the details and the context and when the draft comes back, think about the feedback, try something like:
This reads like a corporate announcement. We're a small team, make it sound like me talking in a team meeting. Also, the key change is buried in the third paragraph, lead with it.
Feedback turns a generic memo into something your team will actually read.
The follow-up that's been sitting in your drafts
A client missed an appointment, a lead went cold, a referral partner hasn't heard from you in a while. You know you need to reach out but the wording feels awkward. Give Claude the situation and ask for a draft. When it comes back, try something like:
This is too stiff. I want it warm but direct, and I don't want to guilt them. Keep it to three sentences max.
One round of feedback turns a task you've been avoiding into a sent message.
The Long Term Effect
These little examples above are ones I've personally experienced and have had clients share with me as well. The reality is these little buggers can eat up a couple hours of your week, so practicing how to work past a first draft is front-loading your time savings in the future.
This is a fundamental skill, don't worry, we'll cover how to automate these sort of actions soon enough.
The Practical Lesson
Nobody good leader hands a task to a new team member and expects perfection on the first attempt. You review their work, tell them what's off and explain what you're looking for. Over time, they start producing work that needs less and less revisions.
Working with Claude is no different. The people getting the best results aren't using secret prompts or a special plan, they're just willing to stay in the conversation a little longer.
Additionally, this habit is just for Claude. Learning to articulate what's wrong with a piece of writing, to say "the structure is right but the tone is off" or "lead with the problem, not the credentials," is a transferrable skill in every working relationship you have.
Don't settle on a first draft.