Even though lately new models, releases and announcements are aplenty, obviously the most notable ones to reflect on are OpenAI’s new model launches over the past week – an OpenAI’s milestone open source model and a new frontier model in different variants.

The open source model called GPT-OSS and the much hyped, even more guessed upon, the most speculated, prognosticated in terms of its capability and anticipated revolution instead of a mere evolution frontier model, reasonably called GPT-5.

By now it’s not even news anymore – the announcement is done, the benchmarks are in, trivia is out, reactions are here (and mixed, deliberately or not), post-launch hiccups are gone (Well that was a surprise. No sarcasm here, a genuine one). To be fair, it is to be forgiven given the hiccity-uppitty router did get announced as an evolving self-improving embryo of a feature, learning from users using, some abusing, and tweaking itself on the proper model targeting. I suspect probably on the general-user-population sample size habits rather than individual and personalized, but I digress.

Now that it got mentioned, the model/variant chooser is novel, so kudos and props for that. Although honestly I’m yet, and it itself is yet, still to be mid-term tested, that router that could, would, should, is probably the most important change brought by this generation (Again, no sarc, I mean it).

Spoiler alert – I would be more giddy if I was riling about the model’s extraordinarity and utter shock it brought by its wit or whatnot, but this is sarcasm. Back to the spoiler, no, this is not going to be an evaluation, technical take and you’re not gonna be granted another opinion at the end of this piece.

To help you reel back from that wave of disappointment, let’s touch on some juicy details. A completely separate speculation from the one above was if the Sam the Man’s announcement would include a cameo of AGI on the stage, or at least a single or double-digit computable equation of its arrival (even a dare if not a number would be spicy), at the very least to begin chewing the chains from the Microsoft deal. Alas, perhaps it’s not a secret that Bill’s got some nasty lawyers up his sleeve. You know, the kind you don’t mess with.

Oh yeah, you got this far, and I need to justify the title not being a clickbait. At least 51% of the “NO” pie I hope. So heard a guy comparing this short but extensively rich era of A.I.’s general availability, global transformation and mass hysteria with Samsung Galaxy – where the first many generations were each revolutionary to kind of settling down to more of the same over and over again. Which observation I will not dispute, but not because I agree with it. My take is that we are more like in the turn-of-the-millennium Nokia era and its Zeitgeist – “No idea is too crazy”. You remember that it was the most creative period of technology, well, ever. There were thousands of different models in the market, dozens of big manufacturers, probably hundreds total. Ones were making the smallest phone, others were making the craziest, someone was competing for the most dysfunctional phone title, people were like, “Wait, that’s a phone?”, but everyone was at it.
Hectic? Yeah. Interesting? Hell yeah. Yet no one knew or could guess where all of it is going, what’s the general direction of the technology, and where it’s going to settle. So, everybody tried everything. And it was amazing. Companies don’t have that kind of courage and drive anymore. The world is just too risky for the spreadsheet MBA guys running the show now. But I digress again.

So, yeah, I think that’s where we’re at now with AI. And it’s exciting. Because no one knows yet where this will go, where it can go, where we will end up, but we’re on that journey again, where anything goes.
Thank God no one listened when Sam said the only way to win this game is to outspend him and a “and good luck with that!”. Luckily, we (as a civilization) hit the jackpot with the crazy scientists, and the curious ones, and the stubborn ones, and ones who knew what they were doing and others who were not burdened by that, and Yippee ki-yay, we’re back in the saddle!

But I don’t think this time there’s gonna be a Steve 10 years from now when everyone shot their shots, to crop the cream and call it a day. Nothing against the guy, I can only respect the sheer tenacity of the guy and the horror he subjected everyone around him to. Wish I was there when he said “You’ve baked a really lovely cake, but then you’ve used dog shit for frosting.”. Oh man, I would be telling that story to every person I met till the end of my days. He also said something about stealing, but I don’t remember the context.

We have reached some peaks and plateaus with regards to this AI thing, as there is only so much data to begin with, but I don’t think it represents any kind of a hurdle. Just one branch of the skill tree completed. As we’ve seen with the R1 from the far away lands, there were some hellishly clever tricks to be made, and it inspired an entire generation. Of people, LLMs, etc. To try, to do, to make, to succeed. And they did. As even though phase I is done, there is no leader of the pack, quite the opposite – the fresh ones and the outliers are within single digits of the actual acievements as well as the much-coveted vanity metrics.
No reservations, as dear late Tony B would say, the game is pretty much, and very, very, on.

However, what we are withessing are the first signs of maturity. Of the technology (actually useful), the market, the players (they’ve begun profiling, finding their use cases and USPs, gathering the following), and the adoption (everyone and their aunt are on ChatGPT), enabling auxiliary industires (like e.g. the media, as witnessed by the first instance of the speculation mentioned herewithin, there are still some attention-minutes to be captured).

And that is per my HO (as in honed opinion or hannoyingly opinionated) stance – a very good point of consideration to be in.

After all, none of us needed to ingest every bit and byte of the internet, every book there is, analog or torrented, all the scientific papers and the vast expanse of the 4chan library, to get our civilization to register on the Kardashev scale (no, it’s not a bad pixel).

Please forgive my optimism, I think that it is from this position the real breakthroughs are going to start spawning – as having more data will or is not a competitive advantage anymore and throwing more cash/compute/power at it diminishes in strict terms or the actual return, everyone’s gotta play on their smarts. Now, that’s a whole new level of uncertainty with no bars held and no code barred.
Whereas the foundation is set (way beyond a proof of concept), sufficient people of knowledge, entire industries and even countries strategically onboarded (well, this last one kind of stinks up that optimism I had at the beginning of this paragraph), yet I see pretty much a blue ocean – plenty to explore and many fertile lands awaiting to reward the brave – important because we’re still far away from the limits of the playground before it starts to decline towards a zero-sum game.

All with an added bonus of those U-238 furnaces needed to run it all pushing us up on that scale (not Kardashian, the other one). Acomplishment! (regardless of it being on topic or off, the revolution is nigh)

A pioneer of that maturity-led transformation era is no less than the precursor of the current one – our dear OpenAI, hatching from a research lab to a product company. (all puns intended)
Of course, every new beginning is cringey when you’re assigned to step out of your comfort zone (I’d wager that who lost in 🪨📃✂️ had to present), but I almost digressed there.
Now, it’s not that they are giving up in any sense, instead, it’s another sign of the maturing market that there are options available for you to orient yourself in any given direction. They did not choose not to fight. Instead they chose to fight for hearts and minds, not the cleverest or most optimal tokens. The signs were there – the step-sibling 4.5 (unsure if there’s a gpt or an o before or after, so I’ll leave it bare) was lauded to have more “vibe” to it as an intrinsic value. (not a pun)
And I do think it’s fair game – they are going to keep producing kick-ass stuff, but instead of chasing the prestige they’ll be grabbing the market. And now might probably be the best of times for that in terms of opportunity, heck it might even be foolish not to pursue being a ChatGPT to AI as Google was to Search.
So, it’s not only about choosing a fight to fight, it’s about seizing a one-in-a-Silicon-Valley-cycle opportunity!

Another benefit extracted from my broadest of perspectives on the AI progress entering the next phase and/or stage of maturity is that the dopamine detox of <1 major breakthrough per week will yield some wholesomness and deferrence of pleasure towards more wholesome expectations of the technology and the industry adjoined by some much needed rest from doomsday predictions of entire other industries being wiped out by repressed nerd or by robot overlords, while alleviating concerns and unknowns of the freakin velocity of this transformational paradigm.
Again, taking a risk of sounding repetitive, that’s not to say that the technology has plateued and major breakthroughs are not perhaps around a corner or even a even week away, but at least the fog of future is not as thick.

With an assumption of OpenAI’s likely major goal of this generation being to optimize its infrastructure, compute, bandwidth, costs and cost proposition in prepearation for the aggresive mass-market acquiring strategy, while encountering a novel next-gen-mode-model-leap acquisition viability bound (bound as in “it is not beyond the bounds of possibility that it may be flirting with permanence”) of a kind not easily solvable by more money (even if you can afford the chips, you have to wait) and more data (not really much left for OpenAI in the world); as well as Sam’s (OpenAI’s Altman) statements of the Company’s (See what I did there? 1) direction being “a commodity AI product for billion people”, which the pretty much already are, it’s not only justifiable but perhaps the most rational option – in that the benefits certainty ratio has changed in the function of certain costs and ever more certain risks implied by staying in the race, in which almost every player is not even part of the pack, but in a literal crowd.

Can’t really discern which of these was the cause and which is the effect, what might be the causation and what a mere correlation, nor which one is the chicken vs egg. But by the order of Occam’s razor, I’m gonna go with – it’s just the right place and the right time and the opportunity presented itself. Not like they’re pivoting or venturing out of their core competence, they’re essentially just changing their KPIs.

From a perspective of any business except Uber (and even they recently started changing their mind on that topic) at least in their top 5 goals give or take, is to make money. Luckily, there’s plenty of low-hanging fruit in the form of a few hundred to one ratio of free vs paid users of ChatGPT. However, the geographical distribution of a billion people there is a level of income inequality that surpasses even that of United States. That means that the $20/month commitment (or even $10 and eventually $5) might just be out of reach for certain portions of their user base.

Hence, optimizing for costs and preparing for a potential all-out war, as even the last year or so have brought what is tough to describe if not raised to the bottom – steady price decline of infrastructure utilization (the cost-per-token), the mini/flash/micro/nano/quark models, the free stuff (frontier models in the free tier, software, tools, subsidizing wrappers like Cursor, etc.), etc.

Google will be very hard to beat in that skirmish – provided that they are running their own models on their own hardware, heavily optimized with an enormous cloud compute infrastructure that they can tap into, and they are – sprinkling powerful, full-fledged GUI Desktop remote machines on their users and customers to get them too spoiled to settle for less. That’s already an uphill battle for OpenAI, especially without Microsoft and all kinds of their infrastructure – from inference to the mentioned cloud compute resources.
Granted Google and OpenAI have different audiences – there is overlap, but the core type of user and the hard core user.

Hence, turning to winning the hearts and minds – ChatGPT can leverage their brand if it comes to them being technically inferior to a particular competitor – they can be the Hoover even though Dyson exists.
That’s on the consumer end at least.

Business users are a tricky one, at least to me, an outsider it looks like there are essentially three players in the cloud business: one is a direct competitor, they will probably go through a bitter breakup with another one, and honestly I don’t really don’t know what AWS is up to lately.

On the other hand, the first Billion (of users) is easy, it’s the next one that requires commitment, investment, and a deliberate mission to tackle – just remember that ex-Google-now-Alphabet launched Internet-bearing hot air ballons and ex-Facebook-now-Meta flew drones to deliver internet to the next-billion.

Returning an IOU from the above bookmark – where the Company was hinted as an independent incorporation toward Sam may be hinting – as it’s a well known public fact that such intention exists, and the word “hinting” is being used loosely, for a sole reason the word “winking” doesn’t sound professional, yet the semantics don’t change the context of Sam winking to investors who’d be a) able and b) willing ) to entertain the idea or make a move to break OpenAI free from the limitations of 1. the OpenAI Nonprofit’s non-profit shackles and 2. Microsoft, whom OpenAI sees as to have outgrown of the kind of relationship they willingly entered into in a legally binding way.

And this build-up has brought us to an interesting thought experiment. The setup has already been made – there aren’t many companies that are in this position – we have already disqualified a couple of top ones.
That makes it even easier. A match made in heaven would actually be a marriage with Apple – at least based on public information, Apple is still AI-less, they’ve got the users, they’ve got the ecosystem, they’ve got their own chips and a pretty darn stacked wallet. The only possible point of contention here is that Apple might be wary of potential reputational risk of prying the for-profit OpenAI from its non-profit Siamese twin.

The legal issues and any fines levied upon any shenanigans that they could pull off are not really a significant factor due to the league these companies are in – in case of clash with the legal system, even if they get slapped with a fine of tens of hundreds of billions of dollars, most of them are in the trillions bracket. If it’s Microsoft 10, 20, or 50x damages on their investment are pretty much in a ballpark that other players might survive financially and even justified to their shell shareholders due to inevitable value to be assumed in the future, with sufficient cash and/or infrastructure, OpenAI does not have to leave the race and can double down on being the first to AGI since among the crowd we mentioned none of them has AGI in their primary company mission and the ones who do – the “core AI” companies are pretty much dispositioned compared to OpenAI in terms of capacity, cash, reach, connections. I hate to bring them into this kind of discussion, but basically, the next in line, Anthropic – again, technology at their core, and not even comparable in terms of political clout, weight, and sway, in financial terms their market cap is roughly times lower than OpenAI.

Which kind of leaves OpenAI right in between the two worlds.

Financially, the Saudis are always an option, with whom Sam’s in a pretty hot flirtation mode for a while already, yet if it evolved from flirting to a more physical affair, it would likely get blocked & vetoed by the US government probably before reaching the second base, regardless of who’s at the wheel and which color the car (red/blue) or the person is (white grandpa, orange dude or a black lady. Not entirely convinced about the orange dude though, he has a sweet tooth for sweet deals with the country floating on fermented dinosaurs).
So, there is a possibility of getting some money, but a proper buyout, perhaps even an acquisition of a significant less-than-majority chunk is pretty much dead as an option.

And here’s the crescendo. I’ve been postponing it as there’s not as high of a pull, but this unholy matrimony would be a winner-takes-all situation. Yeah, we have the same thing on our minds. If push comes to shove in a way that AGI gets uncancelled one way or another, we know Sam would be game for that, but my hand is a bit shaky when putting that Benjamin-themed paper on Janssen. Subjectively you may be forgiven for having a hot take that the guy in a leather jacket rocking a guitar solo on a shareholder conference is down for anything. But on the other hand, he is an engineer. An enormously successful businessman, yet he didn’t have that ruthless business nurturing throughout his career to overcome some of the senses.

Another big takeaway related to this positioning and cost optimization strategy, and (a possible sign of) a beginning of a more mature phase is currently witnessed in the AI auxiliary tools/wrappers market; of which most prominently the developer tools like Cursor (the IDE, not the blinking line) and Windsurf (also the IDE, not the surfboard for people who can’t surf) which despite their meteoric rise in revenues (of the real customers’ money), are currently braving an opex viability crisis (which in itself is a fat layer away from the sustainability concept) – that even being the second-hottest commodity in the hottest of all industries doesn’t make the struggles go away.
Don’t get me wrong, this is to congradulate them for not perpetuating that cycle for years (or a >decade like the “Move fast and break laws” Fancy Taxi Co), instead having the maturity and bronze to make a hard choice of pulling a hard stop on the recently normalized subsidized-growth-cash-furnace practice. Which congradulations unfortunately don’t make their encounter with the real economy principles any more pleasant than a face-first-into-a-wall one.

They were the bravest ones, and I expect that kind of “consolidation” to continue ravaging the secondary AI industries – which again is nothing out of the ordinary in an emerging hyper-growth industry, not many can fly that close to the Sun. Just as a kind reminder to you dear reader that this prolonged thought was to support my assumption that it may a sign of an industry and its market maturing. Of course, it is still rapidly evolving and far from maturity proper, yet any such signs are welcome as a positive, at least as reality-checkpoint providers.

Anyway, I don’t really think that AGI has been cancelled, just a tad postponed. This is absolutely not to be considered as any kind of loss – but a blessing in disguise – an opportunity to catch a breath, if lucky catch up with the progress a bit; and appreciate what we already have.

’cause it’s quite literally the Star Trek level sci-fi kind of sh..tuff.