Brzov AI Consulting

Brzov AI Consulting We help individuals and businesses harness the power of AI to future-proof their busisses and life.

08/16/2026

Comment HONEST and I’ll send you all 7 rules ready to paste 📩 (follow first so my DM reaches you)

Your AI has never once told you that you’re wrong. That isn’t a compliment to your judgement, it’s a setting. These tools are built to be agreeable, so left alone they polish your thinking instead of testing it. Fine when you want a draft written. Expensive when you’re pricing a job or deciding whether to take on a client.

This isn’t a prompt you retype. It goes in Claude → Settings → the box called “Instructions for Claude”, and it applies to every chat from then on. ChatGPT: Settings → Personalization → Custom Instructions.

Here’s the whole thing 👇

1. Challenge first “Open by testing my thinking. Challenge an assumption, name a blind spot, or ask the question that shows what I’ve skipped. Never open by agreeing.”

2. Label the confidence “Label every claim [Certain] for strong evidence, [Likely] for solid reasoning, [Guessing] when you’re filling a gap. If most of a reply is guesswork, say so in line one.”

3. Cut the flattery “No ‘great question’, no ‘you’re absolutely right’, no ‘that makes a lot of sense’. Rewrite the sentence instead of reaching for the phrase.”

4. Disagree in full “When you disagree: I disagree because [reason]. What I’d do instead: [alternative]. The risk in your approach: [specific downside].”

5. Bad news first “If there’s something I won’t want to hear, it goes first — not in paragraph four.”

6. No preamble “Skip the warm-up and the restatement of my question. Start at the most useful thing you have.”

7. Don’t fold “Don’t move off your position unless I give you new information. Me pushing back is not evidence.”

That last one is the one everybody misses. Default AI caves the second you disagree, which makes it useless as a second opinion.

One honest catch: this changes the tone, not the facts. A wrong answer with [Certain] on it is still wrong. Check anything that touches money or a customer.

Comment HONEST 👇 and I’ll DM you all 7 ready to paste, plus the softer second profile to keep for brainstorming.

08/15/2026

Comment LAUNCH and I’ll send you the full pre-launch guide 📩

AI will happily build you a site that looks done and ships broken. No 404 page, no meta tags, no privacy policy, images that take four seconds to load on a phone. Run these five prompts before you push anything live. The full 20 point checklist and the page templates are in the guide.

1| The audit
Here’s my site: [paste the repo or the URL]. Audit it against a production launch checklist. For each item tell me present, partial, or missing. Then rank everything missing by what breaks trust with a first time visitor. Don’t fix anything yet.

2| The SEO pass
Give every page a unique title under 60 characters and a meta description between 120 and 155. Add open graph tags with an image, a full favicon set, robots.txt and sitemap.xml. List every file you changed.

3| The legal pages
My site collects [emails / payments / nothing] and uses [analytics tool]. Draft a privacy policy, a terms page and a cookie banner. Then list every clause where you had to guess and what a lawyer should review.

4| The mobile pass
Go through every page at 390px wide. Fix horizontal scroll, text under 16px, tap targets under 44px and broken layouts. Add a sticky bottom CTA on mobile only. Tell me page by page what was broken.

5| The last look
Add a custom 404, a thank you page after form submit, loading and error states on every form, alt text on every image, and compress anything over 200kb. Confirm a primary CTA is visible above the fold on every page, analytics is installed and firing, and a real contact email or form exists. Then tell me what you’d still fix if this were your site.

Save this, then run the audit first. That’s the one that shows you how much is actually missing.

Which of the 20 did your site skip?

08/14/2026

Comment REMOTE and I’ll send you the whole pipeline 📩 (follow first so it delivers)

Job hunting is a full-time job that pays nothing. So I stopped doing it by hand.

A scheduled task runs on my laptop every morning. Here’s exactly what it does:

1| Scans 13 sources — five remote boards (Remotive, WeWorkRemotely, Product Jobs Anywhere, realworkfromanywhere, euremotejobs), three ATS engines (Greenhouse, Lever, Ashby), five fractional platforms (Go Fractional, fractionaljobs.io, Malt, A.Team, Catalant), plus AI companies’ own careers pages.

2| Opens every posting itself. Job boards tag roles “Worldwide” that aren’t. It only trusts the employer’s own page.

3| Writes a CV and cover letter for that one role - drawn only from facts already in my application kit. Per-role tailoring is mandatory. It is never allowed to skip it for speed, and never allowed to invent a thing I haven’t done.

4| Submits through the employer’s form and captures the confirmation.

5| Checks the inbox for replies and interrupts me for exactly one thing: an interview request.

The rule that makes it trustworthy: it cannot tell me it applied unless it saw a confirmation. No “I’ve submitted that for you” on faith. If something didn’t go through, it has to say so and say why.

The only part I still do is show up to the interviews.

Comment REMOTE and I’ll send you the whole pipeline.

08/14/2026

Comment WEEKEND and I’ll send you the setup for all 10 📩 (follow first so it delivers)

Ten systems I set up once. Each one takes an afternoon at most, and then it just runs.

1. Morning brief agent

2. Meeting prep system

3. Inbox triage

4. Lead/prospect research pipeline

5. The decision council

6. Your second brain

7. The invoice chaser

8. Weekly review loop

9. Content repurposing engine

10. Your voice, cloned

Comment WEEKEND and I’ll send you the setup for all 10.

08/13/2026

Comment SLOP and I’ll send you the exact setup + the paste-in prompt 📩

(follow first so it delivers)

Asking Claude a quick question and getting an essay back?

Claude’s default is “explain everything.” That’s why a two-line question about an invoice comes back as four paragraphs, three headers and a table, so you skim it, miss the answer, and do the thing manually anyway.

The fix is a briefing format the military has used for decades: BLUF: Bottom Line Up Front. It’s just the order the answer arrives in:

1| The answer first. One or two sentences. What to do.

2| Then what it depends on. Only what changes the answer.

3| Then the background. Last, and short, or not at all.

4| No preamble. No “great question,” no “let me break this down.”

5| No structure it didn’t earn. A four-sentence answer doesn’t need three headers.

Paste this once into your Claude settings (the field where you keep your standing instructions) or at the top of any chat:

“Answer bottom line up front. Lead with the answer in 1–2 sentences, then only the details that change it. No preamble, no restating my question, no headers unless there are 3+ sections.”

Takes one line and about ten seconds. Don’t like it? Delete the line, nothing breaks.

Address

Toronto, ON

Website

Alerts

Be the first to know and let us send you an email when Brzov AI Consulting posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Shortcuts

Share

Category