OpenAI's big problem

You’re reading Release Notes, a weekly newsletter covering product lessons from scaling beehiiv and building AI products in the creator economy. (Read the full archive here.)

If you're new here, I'm Jake, a product expert in the creator economy, currently leading new product initiatives at beehiiv.

OpenAI released GPT-5.4 two weeks ago. Supposedly it's their best model yet.

I haven't tried it. And I didn't feel the need to, which is exactly the point.

Six months ago, a launch like this would have set the internet on fire. My Slack channels would be buzzing. WhatsApp groups debating benchmarks. Everyone switching tabs, running tests, sharing takes.

This launch? Nothing. Not a single mention in the Slack channels I'm in. Barely a ripple online. A fraction of the reaction you'd have expected from a release this significant.

The silence is the symptom of a major problem for OpenAI.

Coincidentally, a few weeks before the GPT-5.4 release, I cancelled my ChatGPT subscription. Not in frustration. Not after a bad experience. I cancelled it because it had been completely replaced by Claude & Gemini.

Since then, my LinkedIn feed has been flooded with people making the same decision, and this has only accelerated since the Anthropic Vs. Department of War Vs. ChatGPT fallout that has escalated over the last few weeks. (Read about that here)

To be clear, this isn't a prediction that ChatGPT is dead. It's something more specific: the structural dynamics of the AI market are changing, and people are picking sides (they’re not picking ChatGPT).

This is going to lead to a massive churn problem for OpenAI AND significantly restrict the share of the market they can capture. And unfortunately, this is not the kind of problem they can fix with a better model. GPT-5.4 is proof of that.

Let me explain:

[sponsored] Before I dive into all of the issues that OpenAI is facing, I wanted to share the tool that has made the biggest difference to my AI workflow at beehiiv, Wispr (who were kind enough to sponsor this newsletter).

The best AI outputs come from detailed prompts. But typing long, context-rich prompts is slow - so most people don't bother.

Wispr Flow turns your voice into clean, ready-to-paste text. Speak naturally into ChatGPT, Claude, Cursor, or any AI tool and get polished output without editing. Describe edge cases, explain context, walk through your thinking - all at the speed you talk.

Millions of people use Flow to give AI tools 10x more context in half the time. 89% of messages sent with zero edits.

Works system-wide on Mac, Windows, iPhone, and now Android (free and unlimited on Android during launch).

Try Wispr Flow free →

There is no tool I’ve recommended more to friends/colleagues - trust me, you should give it a go.

Gemini and Copilot aren't winning headlines, but ask anyone at a large non-tech company and they're the only AI tools people have access to at work. They're not winning because they are better models. They're winning because they're already embedded in those organizations. No new login, no new subscription, no sign-off from IT, no perceived risk.

This, combined with the fact these tools are often included for free within the microsoft/google bundle, will ensure massive adoption over time within these companies and create massive switching costs that prevent OpenAI from converting them.

If you need a parallel for how this will probably play out, look at Slack vs. Microsoft Teams. Slack was the first mover, the better product, and the market leader. Then Teams got bundled into Microsoft 365, and adoption quickly caught up to Slack, before leaving it in the dust as it proliferated through the organizations. ChatGPT is about to live through the exact same graph. (TLDR: bundling is hard to beat.)

ChatGPT's strategy was to be everything to everyone. That made sense in 2023 when the goal was mass adoption. It makes less sense now that the market rewards depth over breadth.

Claude picked lanes (coding, writing, data analysis) and those happen to be the lanes that internal champions care about most. Developers and PMs aren't just heavy users and early adopters; they're the ones who have massive influence over which tool the company adopts. When they pick Claude, the organisation follows.

This is exactly how it played out at beehiiv. A few engineers picked up Claude Code, their output jumped (noticeably), others followed. PMs built business-facing use cases, leadership noticed, and full company adoption followed. By the time the decision was formalised, the internal champions were already 100% #team-claude.

Here's the thing that gets underestimated most: what happens after a company goes all-in. Once the team standardised on Claude, the compounding kicked in. Our tech team connected every internal tool (database, backend, CRM etc.), people started sharing workflows weekly, and skills submitted by one person became accessible to everyone. Every bit of incremental usage from a co-worker expands what I can do. Similarly to the microsoft example above, this compounding dynamic creates massive switching costs, which makes life significantly harder for ChatGPT.

This is OpenAI's biggest problem. The "try everything" era is over. Companies are picking a platform, and the two options I keep seeing are: Copilot (if you're already on Microsoft) or Claude (if your technical team has any say). ChatGPT isn't in the conversation the way it used to be.

I get the logic behind OpenAI introducing ads on lower ChatGPT tiers. The math could make sense given that they are under pressure to drive more revenue from chat. But it's going to be terrible for retention, especially when they’re in such a competitive environment. It will likely lead to a downward spiral.

The users most likely on lower tiers are already on the fence; ads give them the nudge to leave (especially when they're starting to get AI for free at work). More churn → more pressure on ads to drive revenue → more ads → more churn. The only exit is a product improvement compelling enough to incentivize users to upgrade, but that means winning back users who've already moved on (good luck).

Even if OpenAI ships something transformative, they still have to solve reactivation, which is a harder problem than acquisition. GPT-5.4 is the clearest illustration: their best model ever, launched to silence. Not because nobody noticed. Because nobody cared enough to act on it.

At beehiiv, we almost certainly won't switch. We've invested real time building workflows, connectors, and team training around Claude. The performance gap that would justify switching doesn't exist.

One of the biggest things that will differentiate company performance and efficiency over the next two years will be what percentage of the workforce can you get using AI consistently. The training alone to get everyone up-skilled on a specific tool and workflow is sufficient to prevent companies even considering a switch.

(Not to mention the loss of momentum from having to re-build skills, connectors, and adjusting workflows etc.)

That disruption immediately eats any small improvement you might've gained from the new model.

None of this means ChatGPT disappears - it has the brand power to still dominate the consumer market for the near future (although I think gemini will catch it within two years).

It just means the gap is widening for OpenAI, and the dynamics driving it aren't in their control. A better model doesn't fix an infrastructure problem. If you asked me to back one horse, it's Anthropic.

Why?

I'm curious whether you're seeing the same consolidation in your own stack, and whether GPT-5.4 moved you at all.

Hit reply and let me know.

Cheers,
Jake

PS. If you’re considering launching your own newsletter, you can get a free month & 20% off by clicking here.