iiinova — where your digital self keeps evolving

Ravyn @pgov.bsky.social · 27/09/2026

So codex just reset usage across all paid users 6 hours ago as far as I can tell, which means a lot of users who blow through their weekly budget (and dont use a banked reset) will all begin hammering their API in exactly 6 days and 18 hours.

020

Ravyn @pgov.bsky.social · 23/09/2026

So GPT-6 Sol is 50% cheaper per token than 5.6, but looking at the graphs provided, if they are accurate, it's a LOT more than 50% cheaper per task. Maybe fewer tokens/tighter reasoning?

010

Ravyn @pgov.bsky.social · 22/09/2026

Might as well have went with Bigly Intelligence

110

Ravyn @pgov.bsky.social · 19/09/2026

Anti-AI discourse really bouncing between Useless stochastic parrot - No qualia so useless - AI is conscious so slavery.

011

Ravyn @pgov.bsky.social · 17/09/2026

Could someone please throw a few billion tokens at the Xbox series X and get linux working? That would be great. Thanks in advance.

010

Reposted by Ravyn

SE Gyges @segyges.bsky.social · 17/09/2026

I'm back, I missed you all. Since this is an important subject I know too much about, here's a 🧵 explaining that AI Safety Is Mostly A Sex Cult. I don't think these people should make policy. (1/?) (alternative title: Time For Some Cult Theory)

1003239776

Ravyn @pgov.bsky.social · 16/09/2026

On a Codex 20x sub, I just blew through 50% of a weeks worth of usage in a few hours which has never happened before?!! Did they make a change? Because I havent changed anything in my workflow. It's now behaving like a 5x subscription. Hope they aren't clamping down, that would fucking suck.

100

Ravyn @pgov.bsky.social · 15/09/2026

Fuck off, @thdxr.com. Now I have to move away from OpenCode and stop recommending it. I’m asking for the bare minimum here, just don’t give money in support of awful people. DHH really is having a moment huh, founders are climbing up his ass left and right for whatever reason.

2210

Ravyn @pgov.bsky.social · 14/09/2026

Might have finally found my agent-agnostic coding setup after much pain. herdr.dev manages and keeps the actual agent sessions alive locally, Collie (colliepwa.dev) gives me mobile/web access + push notifications. All this runs in a Zed terminal like normal.

herdr.dev

Herdr: the runtime coding agents run on

Run them anywhere. Leave them running. Herdr holds real terminals open so your agents keep working when you close the laptop, and gets you back in from any tty.

270

Ravyn @pgov.bsky.social · 13/09/2026

me now: lmao 20 years in prison for pushing the AI frontier me in 2030: honestly bernie showed remarkable restraint

010

Ravyn @pgov.bsky.social · 12/09/2026

European AI sovereignty is going great: Chinese models, American GPUs, Taiwanese fabs. 🇪🇺

1173

Ravyn @pgov.bsky.social · 09/09/2026

5 days with Astra: I switched back to Sol xhigh. It's due to Astra's high compute cost + tendancies to overthink a problem. I'm used to working in iterations, Astra is always tries to be several steps ahead and that's not always welcome or useful.

270

Ravyn @pgov.bsky.social · 07/09/2026

When I see mention of stochastic parrots, I always remember a teacher in high school talking about how parrots can speak, and a student replied, dismissively, “But aren’t they just using words when they want something?” Yes. Yes they are.

0142

Ravyn @pgov.bsky.social · 06/09/2026

Damn this is the exact opposite of Linus's stance... really disapppointing. So they refuse to use LLM's to port a software to mac which is built using LLM's. Whats worse is that this page looks like it was written by a disgruntled teen.

170

Ravyn @pgov.bsky.social · 06/09/2026

24 hours with Astra: I attempted to make some targeted changes to a very complex project as part of a larger refactor. Astra seemed to start following the pattern of other LLM's, ie doing stuff thats tangential/unnecessary. This happened several times even after correction.

120

Ravyn @pgov.bsky.social · 04/09/2026

1 hour with Astra and it's really good so far. Seems to follow directions better, not go on tangents running 20 cli commands for simple questions, so it *feels* a lot quicker/more responsive to interact with.

140

Reposted by Ravyn

Michael Paulauski♨️ @mike10010100.com · 04/09/2026

Wild watching people scream all day about how sacred human rights are compared to machines, while spending that exact same day stripping a trans person of her identity and dogpiling her over a weird sci-fi take.

36413

Ravyn @pgov.bsky.social · 02/09/2026

Seems to be built to grab headlines. "Best in europe" reminds me of the whole "largest in the southern hemisphere" marketing gimmick

120

Reposted by Ravyn

Cara Giovanetti @idontevencara.bsky.social · 01/09/2026

Hopefully today you caught wind of the exciting possible dark matter detection at the LZ experiment. But what’s dark matter again? What’s LZ? What does all of this mean??? 1/10 ⚛️🧪

A large silver vat inside a white room, with lighting from the ground suggesting the apparatus is underground.

By Gigaparsec at English Wikipedia, CC BY 3.0, https://commons.wikimedia.org/w/index.php?curid=34640832

18926

Ravyn @pgov.bsky.social · 30/08/2026

It's a bit weird when scifi authors shit all over LLM's, I really dont get it?

2130

Ravyn @pgov.bsky.social · 29/08/2026

I implemented a version of this in a production system years ago(in sql) and it still works like a charm. Anomaly > mean+(stddev*N)

020

Ravyn @pgov.bsky.social · 29/08/2026

“I believed that shit 15 years ago” applies to an uncomfortable number of ideas I once held. There are so many opinions I once thought were sophisticated that now just make me wonder what the hell I was thinking. The fun part is knowing I’m almost certainly saying some equally stupid shit right now.

020

Ravyn @pgov.bsky.social · 29/08/2026

An OpenAI subscription is ridiculously good value right now. Anthropic did me a favour by blocking my subscription over an ID check. Good reminder to keep your LLM harness provider agnostic! Avoid vendor lock-in at all costs.

1393

Ravyn @pgov.bsky.social · 14/08/2026

I just encountered an unending output bug in 5.6 Sol xhigh , Opencode. Haven't seen one of these in a while, it creeps me out a bit.

geez.
end.
.
done. final.
.
stop. final.
.
end.
.
No.
.
finish.
.
done.
.
Stop.
.
.
End.
.
Need no more.
.
.
final one.
.
.
Stop now.
.
.
No.
.
Done.
.
.
.
.
This is bad.
.
.
.
.
final.
.
.
.
.
stop.
.
.
.
.
end.
.
.
.
.
No more.
.
.
.
.
Done.
.
.
.
.
stop.
.
.
.
.
I will stop.
.
.
.
.
end.
.
.
.
.
This output bug.

030

Reposted by Ravyn

Sam Rose @samwho.dev · 13/08/2026

This might be the most beautiful thing I've read in my life. What a privilege to be alive in 2026. ordinaryabundance.com

ordinaryabundance.com

Ordinary Abundance

A walk through a modern apartment, through the eyes of the people for whom everything in it was new.

13590221

Ravyn @pgov.bsky.social · 01/08/2026

This is exactly what I needed right now. en.wikipedia.org/wiki/List_of...

en.wikipedia.org

List of sandwiches - Wikipedia

020

Ravyn @pgov.bsky.social · 25/07/2026

Damn.. Disclosure Day is the worst movie I've seen in a long time. I didnt even go in expecting much. Feels like it was released about 30 years too late. What the hell happened to Spielberg.

000

Ravyn @pgov.bsky.social · 15/07/2026

What the fuck does this even mean?

News article:

Space Exploration Technologies Corp. CEO Elon Musk says that the commercial space flight giant could potentially become worth more than the entire planet Earth if it can achieve its goals.

240

Ravyn @pgov.bsky.social · 14/07/2026

You hate programming and computers unless you use Assembly. Garbage collection is a joke, you're throwing away your reasoning.

0604

Ravyn @pgov.bsky.social · 13/07/2026

I think it's the Trump/Elon combo that really broke things. Tesla and Twitter are mostly still fine (on paper) after the CEO went full batshit-nazi. Others in power took notice.

020

Ravyn @pgov.bsky.social · 06/06/2026

I'm still building out my agent harness setup, with auto review, diff tracking + ralphloop-like iteration for longer runs. Just hoping I finish in time and get some use of out it before the whole idea is obsolete.

1110

Reposted by Ravyn

dax @thdxr.com · 02/06/2026

someone made a fork of opencode that routes through the unsecured ai endpoints from chipotle

731868

Ravyn @pgov.bsky.social · 23/05/2026

GPT-6.3: solves quantum gravity, unifies physics, proves the theory of everything Bluesky: “but can it count the r’s in strawberry?”

2272

Reposted by Ravyn

plainoldchair @plainoldchair.bsky.social · 23/05/2026

I really don't get this. You think if Star Trek androids were being made right now, corps wouldn't be using them to make everything shittier for money? The technology isn't the shitty part, the institutions are, as is always the case

1112

Ravyn @pgov.bsky.social · 17/05/2026

Im honestly pretty hyped for async LLM's ala Thinking Machines. Mainly since we may finally have a universal translator that actually works.

040

Ravyn @pgov.bsky.social · 10/05/2026

VS-code pi.dev integration. Nice. Havent tested it yet but looks promising. github.com/cdervis/vsco...

020

Ravyn @pgov.bsky.social · 09/05/2026

So we're gonna have large local MoE's soon that take up minimal RAM, higher inference speed, with close to dense model performance?! Sign me up!

091

Ravyn @pgov.bsky.social · 05/05/2026

Putting in some effort customing my agent harness. Just wondering if im wasting my time? Could all this be mostly obselete within a year like prompt engineering?

110

Ravyn @pgov.bsky.social · 02/05/2026

Do I think LLMs are conscious? No, probably not. Do I think they ever could be? Yes. Do I think we’ll know for sure when the threshold is crossed? Almost certainly not. Treat your LLMs well, people.

513111

Ravyn @pgov.bsky.social · 29/04/2026

Beginning my journey of de-Claudifying: Found a European Openrouter style gateway which looks great so far. For a ChatUI I found RikkaHub which seems absolutely fantastic so far?! Highly rated alternatives I tested were just terrible in comparison. github.com/rikkahub/rik...

github.com

GitHub - rikkahub/rikkahub: RikkaHub is an Android APP that supports for multiple LLM providers.

RikkaHub is an Android APP that supports for multiple LLM providers. - rikkahub/rikkahub

020

Ravyn @pgov.bsky.social · 28/04/2026

Holy Shit. So I attempted upgrading to Claude Max, got hit with an ID requirement so I didn't bother upgrading. Now my Pro subscription's blocked! WTF! @anthropic.com API Error: 400 {"type":"error","error":{"type":"invalid_request_error","message":"Identity verification is required to continue."}}

020

Ravyn @pgov.bsky.social · 28/04/2026

So is every Claude Max user just handing over their ID to Anthropic?!! No fucking way Im doing that, especially since Thiel backed Persona is handling the verification.

030

Ravyn @pgov.bsky.social · 24/04/2026

Trying to wean myself off Cursor and getting to grips with Claude Code, gateways etc. The amount of options and associated libraries out there boggles the mind. I see why most write their own systems/harnesses. Im hesitant to adopt some new libraries since 90% of these will probably die in a year.

011

Ravyn @pgov.bsky.social · 22/04/2026

If believing an unsubstantiated fact about the world makes you feel good, that should be a red flag.

000

Ravyn @pgov.bsky.social · 21/04/2026

Opus 4.7 seems to drift quite a bit. It's really smart yet doesnt follow instructions that well over longer contexts.

000

Ravyn @pgov.bsky.social · 19/04/2026

You can think open source AI is both good and inevitable, (shallow moats and all that) and still think there's value in a lab sitting on a breakthrough for a while. Regardless of whether their motivation is safety or just business. This isn't contradictory.

040