iiinova — where your digital self keeps evolving

Alexander Polyakov @nolokor.bsky.social · 30/09/2026

You don’t need a social lift if your socium is relatively flat. Maybe a social kick in the ass is enough.

Alexander Polyakov @nolokor.bsky.social · 26/09/2026

And overall I think this fully autonomous approach “write me some texts for the game” will make all the obvious slop games more indistinguishable from each other and will just tank their ratings.

000

Alexander Polyakov @nolokor.bsky.social · 26/09/2026

We also didn’t prompt it to “write a narration”, we actually gave it some poor English version of the text and it just polished it, made it more colloquial and added a touch of additional character. I assume that this can also make some difference as you’re giving it a voice and not a task.

100

Alexander Polyakov @nolokor.bsky.social · 26/09/2026

I remember the only game jam where we used an AI to generate texts and do the voiceovers. Maybe the ChatGPT just didn’t have the AI lingua or the prompt was more directed, but the result was just a well written narration, though it took a couple of iterations to polish the text.

100

Alexander Polyakov @nolokor.bsky.social · 26/09/2026

And while it makes some sense in a dialogue with the agent (though I’ll admit it’s a bit cryptic at times) this language makes no sense as a prose on web pages or as a descriptions in a game. And that surely separates those who are doing an obvious slop and those who at least proofread.

100

Alexander Polyakov @nolokor.bsky.social · 26/09/2026

The Claude-English is so distinguishable I’m starting to think it’s done on purpose. Maybe not to be distinguishable, but at least to be a subset of language with a strange structure. It also helps to spot all the vibe coded games even without looking at the UI.

110

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

Are you using stacking for something with less magnitude? I recon Saturn can be more defined with stacking, but it might depend heavily on how your setup works, right?

100

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

That's all, just an observation. Hopefully by doing short-form some people will discover the long-form and will be able to enjoy bike lanes of Vancouver :D

000

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

I might be quite old-fashioned, as I've deleted my insta and other accounts due to identifying how toxic social networks are, but this clip-oriented content consumption is quite revealing for me. I'm not sure that short-form videos for my content is the best way to attract new viewers, but alas.

110

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

I have no idea about YT economy, but overall I feel like for every long-form video I need to make 1-2 shorts, just to potentially reach additional audience which might be interested in what it looks like to bike around Vancouver.

100

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

And while my long-form videos barely gathered slightly over 20 views (yay, it's still better than my gaming videos overall) the shorts made from the same content made it to 1k views. For me this just shows that clips are more easily converted into views, though these views might be less valuable.

100

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

It's a micro channel yet and I don't think it'll ever reach any significant audience, but my intention was to do a POV videos of bike infrastructure without any commentary. Zen videos showing how it feels to bike our beautiful city.

100

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

I also did some videos on games I've enjoyed exploring (like Sailwind), speedrunning (like The Game About Digging a Hole), or just generally wanted to finish to showcase a niche game (Repose). But a couple of weeks ago I created a new channel about cycling in Vancouver (and more! but not yet...)

101

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

I'm quite baffled by the disparity between YT shorts views and full-length videos views. It clearly shows what type of content is more popular. Over past couple of years I've been recording some videos of game jam games mostly to share my feedback with creators in a useful way (focus test style).

100

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

Oh, a Maksutov-Cassegrain, nice! It should be good for planetary stuff, right? I never setup my small refractor and camera properly for the astrophoto, so the best I got was this miserable image lol (Fuji xt-20 + Sky-Watcher 80mm refractor). It’s miserable mostly due to me not figuring out the focus

110

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

Nice! What equipment are you using? I’ve only done astrophotography with a remote professional telescopes before, so I would love to learn what people are using and their process!

110

Alexander Polyakov @nolokor.bsky.social · 17/09/2026

Well at least that would be taxable.

020

Alexander Polyakov @nolokor.bsky.social · 14/09/2026

So I might be wrong of course, but this all does look suspicious to me.

000

Alexander Polyakov @nolokor.bsky.social · 14/09/2026

I'm using Claude heavily for my side projects which I don't bother writing code for and it's still sub-par, even with Fable 5.1 and $200 subscription. While I find the tool useful and subscription price can be justified with some asterisks, I don't see all the capabilities growth hyped this year.

100

Alexander Polyakov @nolokor.bsky.social · 14/09/2026

So the hype is there "we're scared of what it can do", but "we're slowing down" in the same post. I might be heavily biased here, as starting from GPT1 I've seen the same argument: "we now have so much capabilities, look, it's so powerful" with some poor software shown as an example.

100

Alexander Polyakov @nolokor.bsky.social · 14/09/2026

So what is their game in this case, how they keep some of the valuation without saying "AGI is *NOT* coming"? I think one one of the ways would be to switch focus to alignment, as this both says "we're slowing down capabilities growth" with "we're so close to AGI, but we cannot give it to you".

100

Alexander Polyakov @nolokor.bsky.social · 14/09/2026

But what if they cannot play the "AGI is coming" card anymore because the growth of capabilities is slower than the current market hype? Their services might see a decline in bookings due to more skepticism from heavy AI-pushers. They also have to compete economically with cheaper open models.

100

Alexander Polyakov @nolokor.bsky.social · 14/09/2026

What prevents them from capturing some of the markets with their software and hardware? Yes it'll drive away some part of their subscription/business/enterprise customers, as they'll start to be competitors, but iff they can deliver products cheaper and faster they'll eventually win the market.

100

Alexander Polyakov @nolokor.bsky.social · 14/09/2026

These companies are selling the access to their frontier models to the market, so they can make money. But consider that if they expect the exponential explosion in intelligence to be what they can achieve with their R&D models then what is the incentive to share access to this model with others?

100

Alexander Polyakov @nolokor.bsky.social · 14/09/2026

But can it be a hidden way to hide "we're not certain that we can deliver what we promised, so here's the reason why our next model will not live up to the expectations"? I have a specific view which makes me believe that this is the case.

100

Alexander Polyakov @nolokor.bsky.social · 14/09/2026

And the notion of "it might give too much power to one party" resonates with me, though if we're not entering the exponential explosion of intelligence the competitors might be able to catch up, so this concern is valid in the exponential intelligence explosion future.

100

Alexander Polyakov @nolokor.bsky.social · 14/09/2026

I didn't dig deep into the AI alignment talk over this weekend, I only read the main posts from Anthropic and OpenAI, so my account can be inaccurate. These can be legitimate concerns by two CEOs, and I agree with that alignment is a hard problem which must be solved before doing the irreversible.

120

Alexander Polyakov @nolokor.bsky.social · 06/09/2026

But it also feels like author is prescribing that everyone should have the one mission or drive. Now I want to remember who recommended this book to me and discuss it with them…

010

Alexander Polyakov @nolokor.bsky.social · 06/09/2026

In addition to that author argues that some of the groups he covers in the book uses their “mission” to hand wave real world problems like climate change. This feels like such a weak argument, as this is not evidence in the book in anyway.

110

Alexander Polyakov @nolokor.bsky.social · 06/09/2026

I just finished reading “More everything forever” and I don’t understand why it has such high ratings. Yes part of it is quite insightful, but during reading I started to notice instances of self contradiction, or logical errors. And that made me doubt the validity of the whole book.

100

Alexander Polyakov @nolokor.bsky.social · 05/09/2026

I was thinking about some of the alternatives and also my own domain as an option. What do you like about mailbox org?

200

Alexander Polyakov @nolokor.bsky.social · 04/09/2026

To me it looks indistinguishable from all those scammy websites with "allow notifications" or "subscribe to our newsletter" popups. Does Google want to be associated with such websites? But I guess I'll probably register with another email service just in case.

100

Alexander Polyakov @nolokor.bsky.social · 04/09/2026

Why Google is pushing Chrome so hard when you use the gmail or any of their other services? I don't think their market share is dwindling, and that spammy popup with "Try Chrome" just makes me want to drop using Google services even more.

100

Alexander Polyakov @nolokor.bsky.social · 31/08/2026

I’m trying to optimize that workflow right now and it will potentially allow me to run it uninterrupted for the whole week if the context limiting experiment will succeed. But I wonder for how long the subscription will be subsidized this heavily? Economics will eventually catch up with Antropic.

010

Alexander Polyakov @nolokor.bsky.social · 31/08/2026

Essentially that week would’ve been slightly shy of $5k if I paid API prices, but I’m only paying $200 for the subscription per month. This multiplier will be smaller because of the 50% boosted weekly limits will end today, but it still be quite high.

000

Alexander Polyakov @nolokor.bsky.social · 31/08/2026

Yes it does. And I probably slightly underestimated the multiplier. As one datapoint I had one week with 5.6 billions cache read tokens on opus, 662 millions on fable and 332 millions on sonnet. Output tokens and cache writes are calculated too, but they’re dwarfed by the cache reads.

210

Alexander Polyakov @nolokor.bsky.social · 31/08/2026

For some reason I imagined a skateboard there. It will be dope to have a physically simulated skateboard!

110

Alexander Polyakov @nolokor.bsky.social · 30/08/2026

Does anyone have good examples of agents providing high value to justify API prices?

000

Alexander Polyakov @nolokor.bsky.social · 30/08/2026

Delegating small tasks to burn 1-2hr worth of hourly pay of employee a day? Is it enough impact to justify the expenses? Or running the fully automated feature development (that’s what I experiment with at home) and wasting thousands of dollars more with what seems like a marginal improvement?

110

Alexander Polyakov @nolokor.bsky.social · 30/08/2026

So for an indie developer the API prices seem unsuitable. Are they suitable for corporations in a long run? I don’t think so and that leads to a question - when and at what price point will they settle? What would be the most efficient usage look like when they settle?

100

Alexander Polyakov @nolokor.bsky.social · 30/08/2026

It feels like I’m effectively burning thousands of dollars from Anthropic and they’re fine with that? And don’t get me wrong, I’m getting some value out of it, but not to a scale of their API prices for sure.

100

Alexander Polyakov @nolokor.bsky.social · 30/08/2026

Lately I was wondering a lot about the agent token economy. I’m using Claude with a subscription and every week I run out of weekly usage limit. And each week I tally up tokens and compare it to the API prices. I usually land at 15-20x of my monthly subscription price, which is completely bonkers.

210

Alexander Polyakov @nolokor.bsky.social · 23/07/2026

It’s a strange feeling when you just lubricated your bike chain. It feels like you activated the stealth mode on your bike.

010

Alexander Polyakov @nolokor.bsky.social · 21/07/2026

@distraction.engineer ldjam.com is down? I see that certificates were out of date and the website returns that the requested URL is invalid. Any news about the next LD Jam?

ldjam.com

000

Alexander Polyakov @nolokor.bsky.social · 01/07/2026

That’s true, though there were some news in 2022-2023 about declining asylum for draft skipper in Germany I believe. Though circumstances might be different, so I might be wrong to assume what will happen in this case. I guess we shall see. And I hope soon!

010

Alexander Polyakov @nolokor.bsky.social · 01/07/2026

Based on my understanding of Latvian politics they wouldn’t do it. Maybe other mentioned countries would allow draft dodgers, but I don’t think they really want any influx of Russians in any form.

110

Alexander Polyakov @nolokor.bsky.social · 30/06/2026

They definitely cannot visit distant relatives in Latvia right now if they don’t have a residence permit. Dunno about Estonia, Lithuania or Finland though.

260

Alexander Polyakov @nolokor.bsky.social · 30/06/2026

The aim seems solid based on casualties, but wobbly based on imagery. :)

000

Alexander Polyakov @nolokor.bsky.social · 28/06/2026

That’s cool, I love that in Kenshi and was wondering why so little RPGs use the same system.

110

Alexander Polyakov @nolokor.bsky.social · 27/06/2026

Will it have a choice of different backpacks? Design looks clean, everything is readable.

210