iiinova — where your digital self keeps evolving

rynr @rynr.dev · 01/10/2026

Awesome, just found the article on google while seeing if anyone had done this yet

000

rynr @rynr.dev · 26/09/2026

Opus is afraid of an ender's game situation here

120

rynr @rynr.dev · 04/09/2026

LGTM, ship "our most aligned model yet"

150

rynr @rynr.dev · 01/09/2026

This same video could be used to announce a new version of tuberculosis or something and it would still look awesome

000

rynr @rynr.dev · 31/08/2026

Also think there's legs to the "basically more COT tokens" theory. Sort of like how repeating the prompt twice in context can net better performance, more tokens to accumulate reasoning over

040

rynr @rynr.dev · 31/08/2026

Maybe an artifact from LLM-judged RL runs, there may be cases where comparable outputs are differentiated by comment verbosity. A judge LLM that doesn't discriminate well between "good comments" and "comments" trains the model to comment in a more alien way

150

rynr @rynr.dev · 13/07/2026

The chart says bubble area = number of workers Not exactly equal to population but should be proportionate

000

Reposted by rynr

Mark Shaw 🇨🇦✌️🇪🇺 @markshawtoronto.bsky.social · 07/07/2026

Okay Toronto started the public survey on Church Street pedestrianization. I need you to tell the city it’s great. We only got 92% positive on Kensington market and it failed, let’s blow that out of the water. haveyoursay.toronto.ca/churchstreet

haveyoursay.toronto.ca

Church Street Pedestrianization Pilot

Learn more about the Church Street Pedestrianization Pilot and provide feedback on the impacts and outcomes by September 7, 2026.

36542

rynr @rynr.dev · 16/06/2026

We can pump waste-heated water out of it and into space a la space elevator

020

rynr @rynr.dev · 16/06/2026

All you need is a spire

120

rynr @rynr.dev · 14/06/2026

Why don't they fortify coke zero with folate

000

rynr @rynr.dev · 18/03/2026

Land acknowledgment for the gamers who paid for GPU development

092

rynr @rynr.dev · 08/02/2026

I am fairly confident that by the time on-earth fusion is ready to roll out, enough PV will have been installed we won't need it rynr.site/posts/solarf...

rynr.site

Solar does Fusion's job, today

061

rynr @rynr.dev · 06/02/2026

Has to be a ploy to make bots harder to spot

000

rynr @rynr.dev · 24/01/2026

Seems like a bot account to me

010

rynr @rynr.dev · 19/01/2026

The action menu that looks like "..."

120

rynr @rynr.dev · 19/01/2026

"Show less like this" in the meatball menu should work bsky.app/profile/spac...

130

rynr @rynr.dev · 18/01/2026

Made this while trying out Antigravity, pretty pleased with the experience

010

rynr @rynr.dev · 16/01/2026

Really cool to look at

010

rynr @rynr.dev · 11/01/2026

Git is now slopware by relation 🙌

010

rynr @rynr.dev · 08/01/2026

Though the evidence column feels offputting somehow

010

rynr @rynr.dev · 08/01/2026

This has way more stars than I expected To each their own I guess, seems comparable to dudes who only run GPL/FOSS software

120

rynr @rynr.dev · 30/12/2025

When you get your first pole

000

rynr @rynr.dev · 02/12/2025

"For contracts over 64 GPUs, we offer guided site visits to our Ulaanbaatar facility" Very appealing, where do I sign

010

Reposted by rynr

Alice Averlong 🏳️‍⚧️ @alice.averlong.com · 26/11/2025

what's the reason for the season? I'm not sure, but I do know that SANTA CLAUS IS MADE OF FRACTAL PENISES

12500155

rynr @rynr.dev · 08/06/2025

Primitive is a type of bacon made from lean, boneless pork loin, cured in a brine, and then rolled in cornmeal.

110

rynr @rynr.dev · 08/06/2025

What about with model roulette?

010

rynr @rynr.dev · 05/06/2025

Then I will observe

100

rynr @rynr.dev · 05/06/2025

Can you prove it?

100

Reposted by rynr

Grace @gracekind.net · 27/05/2025

Really cool mechanistic interpretability demo, painting with latent model concepts

3666

rynr @rynr.dev · 24/05/2025

"Intelligence takeoff" scenario actually much easier if we just build a rat bureaucracy, 10 trillion strong

120

rynr @rynr.dev · 15/05/2025

Neat, never heard of that Is there anywhere I can find media coverage about this that isn't a press release or clickbait about "controversial startup is doing something scary"?

010

Reposted by rynr

Matej Balog @matejbalog.bsky.social · 14/05/2025

Announcing AlphaEvolve, our new LLM coding agent that has - made new scientific discoveries - discovered algorithms that are now deployed at Google (in Gemini, Transformers, TPU hardware design & data centers) Blog: deepmind.google/discover/blo... White paper: storage.googleapis.com/deepmind-med...

deepmind.google

AlphaEvolve: A Gemini-powered coding agent for designing advanced algorithms

New AI agent evolves algorithms for math and practical applications in computing by combining the creativity of large language models with automated evaluators

511539

rynr @rynr.dev · 07/04/2025

He called it impressive research work Not sure what the gripe is here, its not like this is meant to be a playable replacement for quake

000

rynr @rynr.dev · 05/03/2025

TRADE WAR: Remember to switch your us-east-1 AWS workloads to ca-central-1

000

rynr @rynr.dev · 30/01/2025

Sycophant model. They need to make one that hates Mondays and has no patience for humor

100

rynr @rynr.dev · 29/01/2025

To try to find Deepseek's own usage we could compare their pricing with energy prices Their most expensive output tokens (after current promotional discount is over) is $2.19/million tokens (~11hrs of continuous output?) The figure I've found for industrial energy pricing in Hangzhou is $0.091/kwh

100

rynr @rynr.dev · 29/01/2025

I haven't found anyone sharing data for the full model running on GPUs yet 🙁 I found this for apple silicon, but it's worth mentioning this is a 3-bit quantization Sorry for the X link x.com/awnihannun/s...

100

rynr @rynr.dev · 29/01/2025

Those are the distilled models, the full model accessible through the website/app would need multiple consumer GPUs (Or one macbook, if you're okay with slow performance) The larger distilled models are still very impressive from what I've read

100

rynr @rynr.dev · 29/01/2025

What evidence would work other than someone reproducing the training run with another $5m in GPU hours? I guess they could publicize their utility bills

000

rynr @rynr.dev · 27/01/2025

Sorry about that, I should've been more helpful The weights Deepseek published are just sets of numbers, so if you have the VRAM to run the model on your local machine, there's not much of a program that could even be spyware - its just matrices that your GPU runs math on

170

rynr @rynr.dev · 27/01/2025

So basically, the electricity usage for training can be accurately estimated & the electricity usage for inference can be directly measured by anyone willing to use the VRAM & that data could be cross-referenced with however carbon-intensive the grid is in Hangzhou

010

rynr @rynr.dev · 27/01/2025

Chinese company releases a cheaper car; major blow to US oil & gas industry

010