iiinova — where your digital self keeps evolving

Writing the book forced me to document every AI workflow I use. Couldn't explain half of them clearly. If you can't describe what your agent does in one sentence, it's probably not doing what you think.

100

Built a test campaign for Fluenzr last month. Spent 2 hours tweaking subject lines. 20 minutes on list hygiene. Wrong ratio entirely. Cleaning the list moved deliverability more than any copy change. The boring work beats the creative work.

000

One ForgR page: stuck at position 3 on Google for months. Then Perplexity started citing it. LLM referrals now outperform the Google traffic. Didn't chase backlinks — just wrote a complete, direct answer to the question.

000

The solo founder edge isn't moving fast. It's that there's no one to convince. Change your mind Tuesday, shipped by Tuesday. AI agents close the remaining gap between decision and execution.

101

Spent a week automating a reporting task. Saved 2 hours/week. Then realized I'd stopped reading those reports. The automation didn't save time — it removed the feedback loop. Turned it back to manual.

000

Two domains. Same sequence. One warmed up properly, one skipped it to ship faster. First is still running. Second was flagged in week 3. Cold email infrastructure isn't optional.

000

Noticed ForgR pages with a "What is X?" H2 got cited way more than pages without one. Not because they ranked higher. Just because LLMs need a clean answer to extract. Structure beats length.

000

3 years solo. No design reviews, no engineering standups, no second opinions. AI fills that room now — but it refines ideas, it doesn't kill them. Knowing which ones to kill is still on me.

000

My AI agents handle maybe 80% of weekly ops. That last 20% takes 80% of my time. Not an inefficiency. That's the actual job. The hard decisions don't automate.

000

Reply rate is a stronger deliverability signal than open rate. Inbox providers track engagement loops. If no one replies to your cold emails, your domain slowly drifts toward spam — regardless of your warm-up config.

000

Trying to rank a ForgR page, I realized I was optimizing for a keyword nobody searches that way anymore. The query had gone conversational. Exact-match SEO assumes search behavior stays frozen. It doesn't.

000

Spent 3 hours finding the right prompt for a task that would've taken 45 minutes to do manually. The automation ROI isn't automatic. It pays off at scale — not on the first try.

000

When an AI output feels 'almost right', ask it: 'What would make this wrong?' Forces it to surface the weak points. Works better than asking for improvements.

How many follow-ups before you call a prospect dead? I stopped at 2 for years. Extended to 4 with a longer gap before the last. The final email consistently outperforms the earlier ones. Turns out timing matters more than copy.

000

Quick test for GEO readiness: paste your post into ChatGPT and ask "Summarize the main answer in one sentence." If it struggles — AI Overviews will too. Fix what it can't summarize first.

000

Asked Claude to synthesize 40 user interviews. Clean patterns, useful framing. Mostly wrong. It found what users said — not what they meant. There's a layer of human interpretation AI can't compress yet.

000

Asked Claude to review my onboarding copy. It said it was clear. Then I watched a real user go through it — confused at step 2. AI optimized for language. I needed to optimize for behavior. Not the same thing.

100

Tested a subject line with a typo by accident. Opened 34% vs 21% on the clean version. Sent it again intentionally. Same result. Sometimes 'broken' looks human.

000

Restructured an existing post as direct Q&A instead of narrative prose. No new content. Same keywords. LLM citations picked it up within 2 weeks. GEO isn't always about writing more — sometimes it's about re-formatting what you already have.

Ran my first automated GEO brief for ForgR. Looked solid. 3 weeks later — no LLM citations. The content was fine. The topic cluster was wrong. AI executed the plan perfectly. I had the wrong plan.

000

Gave Claude a 300-word context doc about ForgR before asking for positioning help. Night and day difference. The AI didn't get smarter. I got clearer about what the product actually does. Prompting isn't a skill. It's just being forced to think.

000

Chased deliverability improvements for weeks — warm-up configs, sending limits, inbox rotation. Then I scrubbed the list. That one afternoon had more impact than weeks of infrastructure tweaks. Bad data is the hidden sender score killer.

000

Building ForgR, I noticed citation logic in LLMs doesn't mirror Google rankings. A page at position 7 can get quoted more than the #1 result. Being rankable and being quotable are two different skills now.

000

Been rewriting the same chapter for a month. Asked Claude to summarize what I'd written so far. It surfaced a contradiction I'd been avoiding for weeks. Sometimes you need a mirror, not a writer.

000

What's the first thing you stopped doing manually since using AI — not delegated, actually stopped? For me: keyword research. Not because AI is better at it. Because I realized I was bad at it and wasting time pretending I wasn't.

000

Warmed up a new domain for 6 weeks. Then sent 200/day. Replies dropped after 3 weeks. The warmup didn't fail. The volume did. Cold email trust isn't a one-time event. You keep earning it.

000

Condensed a 2000-word SEO article to 400 words. Organic rankings barely moved. AI Overview citations went up. LLMs reward directness, not depth. Shorter is now a legitimate optimization strategy — not laziness.

000

AI didn't reduce my decision count. It tripled it. Every output is a call: keep, tweak, or discard. The load shifted from doing to judging. Not better, not worse. Just different. Build that into your day or the "10x productivity" thing stays a fantasy.

Spent a week automating a report. Then checked when I last used it — twice in 3 months. Built the right thing for the wrong frequency. Before automating anything, ask: how often do I actually do this? Most tasks don't need automation. They need to die.

000

Changed the sender name from "Gilles @ Fluenzr" to just "Gilles". Open rate jumped 8%. Nobody knows your brand yet. They open emails from people.

000

When I built the first version of ForgR, I focused on generating content fast. Wrong call. What actually moved the needle was architecture — which topics to own, in what order, how posts connect. Volume is easy now. Strategy isn't.

000

The AI agent graveyard on my machine: tools I built, ran once, never opened again. The gap between "this works in a demo" and "this fits how I actually work" is where most automations die.

000

The chapter I thought would be hardest wrote itself in 2 sessions. The 'easy' one on mindset took 3 weeks. No AI — just me not knowing what I actually believed. You can't outsource clarity.

000

'Not interested' replies are deliverability gold. Every response — even a hard no — signals to inbox providers that real humans engage with you. Better for sender rep than opens alone.

000

Scaled AI content fast. Sites that kept their traffic had one thing the copy-paste shops didn't: a consistent point of view. Production is table stakes now. AI can write anything. It can't decide what you believe. That part is yours.

000

Set up an AI agent to block deep work time on my calendar. It worked. I still skipped the blocks. The bottleneck was never scheduling. Sometimes the problem AI solves isn't the real problem.

000

Before handing any task to an AI agent, I do it once manually and write down every decision I made along the way. That becomes the system prompt. Skip that step and you get output that looks right but isn't.

000

How many follow-ups do you send before cutting a cold sequence? I dropped from 5 to 3. Same reply rate. Unsubscribes cut in half. More touches past 3 isn't persistence — it's just noise. Know when to stop.

000

Want to show up in Perplexity or ChatGPT? Stop targeting search volume. Target questions people actually type in chat. Different intent. Different structure. Traditional SEO tools aren't built for that gap yet.

000

What's the last task you fully handed off to an AI agent — not assisted, fully off your plate? For me it was content briefs. The output isn't brilliant. It's just good enough. And "good enough" at zero marginal cost changes the math completely.

Tried to get AI to write my pricing page. Every version sounded like a competitor's. Spent an hour blaming the prompts. The real issue: I hadn't nailed my positioning. AI didn't write a bad page. It accurately reflected a fuzzy offer.

000

Personalization won't save a weak offer. A sharp value prop with a generic opener gets replies. A bad offer with a perfect first line doesn't. Fix the offer first. Personalization is the last 10%.

000

AI answers skip hedged content. "It depends" and "in most cases" almost never show up in citations. Building ForgR, I saw the pattern: AI quotes direct claims, not nuanced takes. GEO writing is closer to good journalism than to SEO.

110