Hey there,
I pay $100 a month for Claude on a Max plan. And last week I hit my limit anyway.
I wasn’t even doing doing anything heavy, but I hit it because:
a) Claude is struggling to handle the inflow of new customers so they started limiting tokens every 5-hour session
and b) because I am a pretty MESSY user.
I have seven connectors ON when I needed one. Running Opus model on tasks Sonnet could handle. Uploading raw PDFs instead of converting them first to .md files.
In other words, I was paying the "lazy tax."
And I think a lot of you are too.
This came up in my membership a couple of weeks ago. One of the members was frustrated about suddenly hitting limits, and the thread that followed turned into a collective tips and tricks session.
Anthropic has admitted they're at high capacity right now, so limits are tighter across the board. And we, Claude users now have to figure out and rethink how we use it. 🙄
So I took everything we were sharing in the membership, added my own research and testing, and turned it into 20 tips for respecting your tokens.
Full breakdown on YouTube this week or check out this interactive resource I created instead.
Three of them hurt the most to realize:
Stop defaulting to Opus. It burns tokens roughly 5x faster than Sonnet. For most work (content, rewrites, brainstorming, planning), Sonnet is more than enough.
Audit your connectors. If you've got eight things connected in Claude Cowork, Claude checks every one of them on every prompt before doing the real work. Type
/contextin any chat to see exactly where your tokens are going. When I did it, my MCP tools/ connectors were in my top three.New task, new chat. Long threads feel efficient. They're the opposite. Claude re-reads the whole history every send, outputs get worse the longer you go, and you pay more for worse answers.
The other 17 tips are in the video - including converting files to markdown, editing instead of follow-ups, tricking the 5-hour window, and using projects properly.
Two things I'm paying attention this week
Claude Opus 4.7 dropped - and the reactions are mixed
Anthropic released Opus 4.7 yesterday. On paper it looks like an upgrade. In practice, the community reaction is more divided than I expected.
What's new:
It follows instructions more literally. This one I've already felt in my own testing. Prompts I wrote for 4.6 sometimes produce different outputs now, because 4.7 doesn't skim or skip parts the way earlier models did.
Big jump on multi-step agentic work. Anthropic reports a 14% improvement on complex multi-step workflows, roughly a third as many tool errors, and better "loop resistance" (meaning it doesn't get stuck repeating itself). For anyone running agents, this matters.
Much better vision. Image resolution more than tripled. It can now actually read dense charts, diagrams, fine print on documents, and screenshots without hallucinating. On the visual-acuity benchmark, it went from 54% (4.6) to 98.5%.
❗️The part you need to know:
Pricing technically stayed the same, but the tokenizer changed. This means that the same prompt can use up to 35% more tokens in 4.7 than in 4.6.
Not great news. As someone wrote on Reddit: “just by thinking about Opus 4.7 i have exceeded my limit“ 😂
My take: it's too early to tell if the quality gains are worth the real-world cost bump.
I'm cautiously optimistic on the instruction-following and multi-step improvements (those are the two things I actually need for the workflows I run).
But I'm also not rushing anyone to switch.
If 4.6 is working for you, let it work.
I'll report back in a couple of weeks with what I'm actually seeing on my own usage.
Canva AI 2.0 is here and its BIG news!
This one I think is genuinely underrated.
Canva just made the move everyone building with AI has been waiting for.
Canva is no longer a design tool - it’s repositioning itself as an agentic workspace.
The company called this its biggest shift since 2013 (when it moved design from desktop software into the browser).
They've rebuilt the platform around AI from the ground up, using their own foundation model (the "Canva Design Model") trained specifically for design workflows.
Quick context: for two years now, those of us creating digital products, social content, carousels, lead magnets, and sales pages with AI have been stuck with the same problem.
We can generate the copy, the outline, the strategy, even the code but the moment we need something designed, we're jumping out of AI back into Canva to manually pull it together (unless you are using Canva Connector in Claude).
Every LLM has felt like it was missing that Canva design layer to make things look pretty 🙂 Canva 2.0 is taking steps to keep you in Canva from the get go.
What changed:
Conversational design. You describe what you want in natural language - a carousel, a pitch deck, a launch graphic and it builds it from scratch.
Object-based outputs, not flat images. Everything it generates is made of individual editable objects. You can swap one headline, one image, one font, without regenerating the whole design.
Persistent memory for your brand. It learns your style, colors, fonts, tone over time and applies them automatically.
Connectors to Gmail, Notion, Slack, Zoom, Google Drive, and Calendar. It can pull from your actual business context - emails, meeting transcripts, calendar to build briefing docs, sales pitches, newsletters, decks.
Canva Code 2.0 with HTML import, which quietly opens up a lot for people building landing pages and web assets and dipping their tows in vibe coding.
Why this matters for us specifically: This could be a HUGE leap for digital product creation!
You can potentially now create digital products more intuitively, give it access to your input (long form content, old templates, documents, client notes etc )and it will create a workbook or any other digital product in your brand colors in seconds.
I will be testing this over the next week and will report back!
Have a great weekend!
Until next week,
Your AI Solopreneur Bestie,
Elena
If you want to explore how to actually use AI for your business with 200 + other solopreneurs, the AI Solopreneur Club is where the magic happens ✨
