At MetaCTO, we build custom mobile apps that help businesses grow new revenue streams, create online communities, increasing customer LTV with digital products, and make client communication more personal.

Christopher Fitkin

“AI makes devs 10x more efficient.”

BS.

Everyone’s acting like AI “unlocked” some hidden 10x layer of performance. But that’s not what’s happening in the real-world.

Allow me to rant…

- Realistically, most devs aren’t trying to be 10x better. They just want to finish their tickets with less effort. Then go home & spend time with their family for whatever reason (joking 🙃 )

- The hardest-working engineers are now buried under waves of low-context, AI-generated code. Good chance they’ll burn out & quit. And good luck with the juniors, because…

- Junior engineers are skipping the painful learning phase. They generate solutions instantly… but ask them to explain why a solution works (or doesn’t)… 🦗🦗

- Long-term cost of sloppy abstractions might not show up this quarter. But tend to show up 18 months from now when everything is harder to change & nobody remembers why it was built that way

- Even if features ship faster, you’re still bottlenecked by meetings, alignment, approvals, compliance, legacy systems, etc. Code was never the only constraint - which brings me to…

- Companies do NOT have a code shortage. Nor do they have a shortage of ideas… just a shortage of great ideas. When building was expensive, that friction acted as a filter (saving companies tons). Now more mediocre ideas get built… and costs tick up, up, up

If you don’t fix idea quality, incentives, and decision-making first… AI’s not making your dev-team 10x superhumans.

It’s simply accelerating what’s already broken.

Did Anthropic just kill the “AI makes developers 10x more productive” narrative? Their new study was… interesting.

The last few years, we’ve been told by well-known companies and execs that if you’re not using AI for coding, you’ll be left behind. That productivity is 10x. Maybe 100x. That software engineering as we know it is basically over.

But a recent Anthropic study showed much more nuance. (yes, that Anthropic… the AI company)

In their randomized study, developers using AI completed tasks faster in some cases - but overall there was NO significant average speed gain in the main study.

What caught my attention was that developers who relied heavily on AI performed worse on conceptual understanding, code reading, and debugging. On a 27-point knowledge quiz, the AI group scored 17% lower.

From my experience leading dev teams, this tracks.

The real productivity gains I see from AI is scaffolding, syntax help, quick exploration of edge cases, and reducing context-switching friction. But if a developer stops understanding the system they’re building, that debt shows up later… in debugging, architecture decisions, and long-term maintainability.

I’m not anti-AI. We use it every day MetaCTO.

But if your strategy is “just let the model write it,” you might ship faster this sprint… and pay for it in comprehension, skill development, and future velocity.

As always, the hype cycle is louder than the data.

Something weird is happening… Software engineers were supposed to “die off.” Instead, they’re getting more expensive.

Anthropic is paying senior engineers a median of $570k (with some roles pushing much higher than that).

But at the same time… Anthropic CEO, Dario Amodei, is on stage in Davos saying AI can handle most coding in 6–12 months.

So which is it?

What I think will actually happen… the job won’t disappear, simply mutate.

These engineers aren’t grinding out loops all day. They’re deciding what AI output is safe to ship. They’re designing system boundaries, & thinking through scale, failure modes, security, edge cases.

They’re the ones who get buzzed when something breaks at 2am. (And well-paid for it!)

My final thought… AI isn’t removing engineering as much as it raising the leverage (and cost) of the people who understand the whole system.

“Engineering is dead” makes a click-baity headline.

But the people who can think architecturally are becoming more valuable, not less.

First it was vibe coding.

Now, I’m seeing “vibe engineering” everywhere.

We’ve all heard of “vibe coding” - the fast, loose, AI-prompt-driven approach that skips silly things like ‘actual expertise’ & ‘can actually work’. Instead, it just… hopes for the best. 🤷‍♂️

But “vibe engineering” is the opposite side of the spectrum.

From my understanding… it’s when experienced developers use AI tools intentionally to accelerate production-quality work (while staying fully accountable for the outcome - crazy, I know).

I like that idea.
I think it captures where we actually are right now.

We’re at the very beginning of learning what we can achieve with these new tools. There’s no perfect playbook. We’re still experimenting with what to build, how to collaborate with agents, and where to draw that fine line between automation and craft.

And that means we’re going to get some of it wrong. That’s fine. The learning curve is part of the job.

If anything, this should be the most fun phase in tech - when the rules aren’t settled yet. We get to play, explore, and figure out what “vibe engineering” really looks like in the real world.

So maybe don’t take it all too seriously just yet.

Experiment. Break things. Learn what works. That’s what a good engineer does anyways.

PS:
We might need a better name than “vibe engineering,” though. Not sure how much more “vibes” I can take 😅

New AI security report out & it’s BAD news for Claude, Codex, Cursor, Replit, etc (or at least, cause for major concern)…

Researchers at Tenzai tested the most popular AI coding tools by having them generate common apps like shops and forums.

Result:
Every. Single. One. shipped with real security flaws.

And this wasn’t just theoretical… they discovered hackers are ALREADY exploiting:

- Permission bugs exposing other users’ transaction data
- Checkout flows that let users enter negative quantities and get paid
- Business logic that allow unlimited refunds, discounts, or credits (bc edge cases weren’t properly defined)

AI writes code fast → yes
AI doesn’t always understand incentives, abuse, or edge cases → also yes

My advice if you’re building anything relating to money, users, or permissions (i.e. everyone)…

- Slow down just a bit
- Reviews, testing, checks & balances are your friends
- Don’t confuse “it works” with “it’s safe”

Speed is a weapon. Unreviewed speed is a liability.

undefined undefined

AWS Certified Developer - Associate

AWS Certified Solutions Architect - Associate

Partner at MetaCTO · Custom Mobile Apps 📱

AI to maximize revenue through custom digital strategies. nova identifies marketing opportunities & creates strategic action plans to unlock business growth.

nova

Power Digital’s SPRnova technology platform helps companies evaluate and analyze first-party data, which businesses collect directly from their customers and own themselves, and then helps in assessing the data and developing marketing and growth strategies based on that analysis.

Court Square Capital Backs Marketing Technology Company Power Digital

Kite will prioritize recommendations in order of highest impact from our 100,000+ campaign experienced data sets. You can have the best search performance in the world, but how do you scale other...

meetkite.com

Tenzai

Technology Founder & Fractional CTO

WILD (but unsurprising) stat of the day: 56% of chief execs report NEITHER increased revenue NOR lower costs from AI over the past 12 months.

(from PwC’s 29th Global CEO Survey)

I see it everyday… companies adopting AI because someone above them said, “We need to be doing something with AI.”

But friends… that’s not a business strategy!

AI only works when it’s the best tool for a clearly defined job:

- Reduce time spent on X
- Increase output of Y
- Improve quality or consistency of Z

Track & measure.

If good results → keep 
If bad result/no improvement → dump

That’s it!

PS: Just for clarity… 30% of chief execs reported higher revenues linked to AI over the past year and 26% cited lower costs. So if you’re doing it right… more power to you! I’m just seeing a lot of people NOT implementing AI right - data seems to back me up.

Founder: “Are we over-hiring?”

Me: “That’s the wrong first question.”

The better one is: “What behavior is our hiring strategy incentivizing?”

If you bring in more engineers than the roadmap requires, and there isn’t a clear mandate tied to outcomes, something predictable happens. Work expands to fill the sprint. Tasks get stretched. Finishing early becomes risky.

The incentive structure shifts from “What can I do to make the most impact today?” to “Oh, I should look busy!”

And once that mindset creeps in, it’s hard to undo.

On the other side, I’m also seeing some companies freeze hiring because they assume AI will magically double output. But if you don’t redesign expectations, ownership, and accountability… adding AI just layers confusion on top of an unclear structure.

Headcount sends a message.

It tells your team whether you value speed, experimentation, stability, or optics. But most importantly… what their role is!

If you’re hiring (or not hiring) without a clearly defined 12–24 month direction and measurable outcomes tied to each team, you’re not just risking inefficiency.

You’re training your org to optimize for the wrong thing!

My advice to mobile app startups…

Ship early → Learn fast → Stop being wrong sooner.

I’ve helped building a ton of apps in the last 20+ years. And what I’ve noticed is early-stage app growth is nothing more than a race to certainty.

You don’t get certainty from planning meetings, sign-offs, or yet another round of “let’s polish it.” You get it from shipping something real and watching real users do real things.

That’s it!

Most founders wait because they want confidence. The irony is waiting is what kills confidence. Every extra week you spend perfecting the wrong thing just burns runway and delays the only feedback that matters → user behavior.

My rule is simple…

If the cost of being wrong is reversible and contained, ship it.
If it’s irreversible or it damages trust, slow down and do it properly.

That’s the whole game.

Shipping early doesn’t mean shipping garbage. It means you’re ruthless about scope and honest about what you’re actually trying to learn. If your big bet depends on one assumption, don’t build the full system first. Build the smallest version that tests that assumption without embarrassing you or breaking the product.

And when something misses? Kill it quickly.

The best part… speed compounds. Because your team is building a habit of momentum. You ship, you learn, you adjust. That loop becomes your advantage.

Now, I feel the need to say this or I’ll annoy someone I’m sure… YES, obviosuly be selective about where you slow down - payments, data privacy, anything that can corrupt data, lose money, or make users feel unsafe. Those are one-way doors. Treat them like it.

Everything else?

Ship it, measure it, talk to users, and iterate like your runway depends on it (because it does).

My recent post about “tech debt” blew up.

Which tells me something…

A lot of companies are drowning in it. So I thought I’d give my thoughts on ‘why’.

I don’t believe tech debt comes from “bad developers.” It comes from decisions made under pressure: deadlines, investor expectations, feature requests, growth targets. And yes - sometimes from developers who just wanted to ship and move on.

—
In early-stage startups, tech debt is often intentional.

You should take shortcuts when speed matters more than elegance. The problem isn’t the shortcut. The problem is pretending it didn’t happen.

—
In bigger companies, it’s usually different.
The system works… but it’s fragile. Slow. Painful to change. And management doesn’t see the cost because the compiler doesn’t complain. “It works” - until velocity drops, bugs increase, and your best engineers start pulling out their already thinning hair.

If your product is unreliable, that’s not tech debt. That’s a defect (i.e. every change feels like brain surgery).

And the worst way to fix it = announcing a “tech debt sprint” (with a massive rewrite). That’s usually emotional, expensive, and disconnected from real business value.

The only approach that consistently works is boring.

Every feature leaves the codebase slightly better than it was.

You add logging while you’re in there. 
You clean up the naming. 
You remove the weird conditional you no longer need. You introduce an abstraction instead of stacking another hack on top.

Because the reality is speed creates tech debt. But lack of discipline lets it compound.

If your team treats tech debt as “someone else’s future problem,” eventually it becomes everyone’s present crisis.

Anyone know a good early career Product Manager? Asking for a friend.

If you hear: “Let’s appoint an AI champion.”

👆 99.999% fail rate.

Okay, let’s assume the best… the person is smart, motivated, capable.

But if they’re handed this golden title and responsibility… but don’t have AUTHORITY to actually do/change/update anything… well, it’s doomed.

Because what happens is pretty predictable. The “AI champion” starts making suggestions, teams push back, and suddenly every decision turns into a debate. They’re stuck convincing, not deciding. Over time, progress slows, politics creep in, and eventually everyone wonders why “AI isn’t working for us.”

And your ‘AI champion’ has now become your ‘AI fall guy’.

Eventually decision-makers assume the problem is culture, training, or effort. But it’s not. It’s ownership. If AI is meant to be strategic, it can’t be optional.

If your ‘AI Champion’ doesn’t have real backing from leadership, you don’t have AI leadership at all. You’ve got a future punching bag.

Authority first.
Adoption comes after.

Something big is happening in marketing.

Traffic doesn’t just come from Google anymore. It’s coming from ChatGPT. Gemini. Claude. And most of it looks like “direct” or “organic.”

That’s new.

We’re entering a world where AI is becoming the first touchpoint. Users ask a question. The model gives a recommendation. They click.

Even more new… these users often convert better. People think a conversation with their friend, ChatGPT, feels like a recommendation from a trusted source… not an ad.

But there’s a catch…

You can’t rank-check ChatGPT. You can’t reliably see when you’re cited. Sometimes you’re paraphrased without a link. Sometimes your traffic shows up unattributed.

We're seeing traditional attribution break in real-time.
So what do you do?

FIRST, accept that LLMs are a new organic channel. Treat them like early SEO… a little confusing but high leverage.

SECOND, write content that AI can quote cleanly… clear answers, structured summaries, comparison tables, FAQs, content that stands alone out of context.

(We’ve done this @MetaCTO and in Jan. 2026, we averaged 100+ organic visits/day - 277 peak. And climbing!)

THIRD, instrument aggressively… use UTMs anywhere links might be picked up, implement deep linking properly, track web-to-app flows, and use schema markup so models understand and cite you correctly.

My final thought to you… don’t “optimize for AI” & pray for the best.

Focus your energy on building measurement into every AI-touchpoint before the channel fully matures!

It’s 2026. And companies are starting to ask a valid question: “Where are the AI gains?”

Recently I was scrolling through a Reddit board when I saw a webdev raise an interesting point…

”I'm only one dot on the chart but my workplace has whatever license offers Copilot and I know no one on the dev, marketing, or design teams use it.”

Not surprising.

Especially considering PwC’s 29th Global CEO Survey came out a few weeks ago and 56% of chief execs said they had NOT seen increased revenue or lower costs from AI (over the past 12 months.)

I’ve heard the same from folks I’ve spoken to.

From where I sit, most large orgs are still VERY EARLY on the AI maturity curve. They bought the tools, approved the spend, announced the initiative… But they never changed workflows.

So nothing changed.

For us MetaCTO, the gains have come when we made it a crucial, mandatory part of our workflow (and made sure the team was aligned).

Then we measure, constantly. Cycle time. Bug rates. Rework.

If it doesn’t improve something measurable, we adjust.

And if it still doesn’t help… we don’t use it!

(Image: PwC’s 29th Global CEO Survey)

The MetaCTO team is celebrating today because our Firebase Authentication Plugin has been ACCEPTED on the Strapi Marketplace! 🎉

This plugin makes it easy to use Firebase Authentication with Strapi v5 without building your own custom auth layer.

Or more simply… your users authenticate through Firebase (Google, Apple, Phone, Email, Magic Link, etc.) and the plugin handles the rest.

It verifies the Firebase token, automatically creates or syncs the user in Strapi, and returns a Strapi JWT so your app can securely access content.

Few highlights:

- Multi-provider authentication (Google, Apple, Email, Phone, Magic Link)
- Automatic user sync between Firebase and Strapi
- Support for phone-only authentication flows
- Encrypted configuration for enterprise security
- Built-in activity logging and admin tools

Huge shoutout to the team who built and shipped this! Getting accepted into the marketplace is always a great milestone.

If you're building with Strapi + Firebase, feel free to check it out.

I’ll drop the link in the comments!

Anyone in digital marketing looking for an opportunity to grow and work with some of the best people I know?

COO @ Pesty Marketing | Operations, Systems, and Software for a High-Growth Vertical Agency

Announcement time 🎉

I'm the new COO at Pesty Marketing.

For the last two years, I ran my own consulting company. Worked with great clients. Helped leaders and teams grow. Built training and operational programs from scratch. It was fulfilling work.

But one thing was missing. Being part of a team again.

I thrive in startups. I thrive in the pace of agency life. Being inside the machine, not adjacent to it. Helping clients hit their goals and watching it compound.

And that's what I get to do again.

Five months in. The growth is real. We've already had some of the biggest months in company history. And we're not slowing down.

And what we're building is different.

We're not just a "marketing agency".

We're pushing hard into AI, automation, and reporting systems that connect inputs to outcomes. We're using Claude Code to build custom NPS and CSAT software. Internal reporting dashboards that give us real visibility.

Shoutout to my CEO, Andrew Peluso, who has a clear vision and gives me the runway to execute. Thank you.

PS... we're hiring 😉

DM me or check the links in comments.

If you're thinking about shipping a vibe-coded app into production... you should talk to Jamie Schiesel first. 🤖🚀

Technology Leader | Driving Innovation | AI Enablement

We just completed a technical audit on a vibe-coded app. 🤖 

It was production-ready... except it wasn't.

Tools like Replit and Lovable are genuinely impressive. Apps are feature-rich, well-organized, and have excellent documentation. (better than most human-written codebases)

But when we dig deeper, it falls apart fast.

What we found:
  - The database was wide open. All access control lived in JavaScript not at the database level. Any authenticated user could escalate their own permissions, create fake purchase records, or modify subscription tiers. 
  - Missed/Inconsistent Endpoint Authentication: One was an open email relay... anyone with the URL could send emails from their domain. The other could trigger mass data deletion across a subset of critical tables. 
  - Lacked Feature Completness: Many features were half baked, leading to bugs, dead ends, and a poor user experience overall. 


AI moves fast. It optimizes for "does it work?" not "is it complete?" or "is it secure?" Without a human in the loop asking the hard questions, the gaps compound quietly, threatening scalability and exposing businesses to real security risks.

Vibe coding is a powerful starting point. It's not a production strategy.

Lesson here. If you're thinking about taking an AI-generated app live, get a technical audit first. 


#AI #SoftwareDevelopment #TechLeadership #VibeCoding #AIEngineering

Dream of selling your app?

This is what buyers tell me they’re looking for…

They want confidence.

In particular, these signals below consistently raise valuations (and make deals move fast):

- Retention > everything

If users stick, buyers assume PMF is real and revenue is predictable. If users churn fast, nothing else matters.


- Clean numbers

Most deals start with a multiple on EBITDA or MRR, then move up/down based on risk. Messy financials = they assume the worst.


- Low operational burden

If your app needs you personally to keep it alive, that’s not an asset… it’s a job. Buyers like apps that run clean: stable infrastructure, documented basics, predictable costs.


- A believable growth story

This is actually underrated, in my opinion. A simple narrative: what’s working, why it’s working, what you tried that didn’t, and what the next obvious levers are.


- Traffic quality (and diversity)

Some buyers love organic. Some love paid. Almost everyone hates “one channel or die.” A mixed engine reads as durable.


If you want a higher valuation, your best move (usually) isn’t adding features. But tightening the boring stuff: retention, reporting, ops, and the story.

I hate to say it’s that simple… but it can really be that simple.

MetaCTO

Back-end Developer

Integrated Solutions For Ports (ISFP)

Software Developer

A good programmer is someone who always looks both ways before crossing a one-way street." ~ Doug Linder

taplio

The Battle

Christopher Fitkin

Omar Akram

Followers

Posts

Engagements

Likes

Performance ScoreChristopher Fitkin's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Christopher Fitkin

Omar Akram

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Christopher Fitkin's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreChristopher Fitkin's LinkedIn Performance