iiinova — where your digital self keeps evolving

index @indexx.dev · 26/09/2026

satan remarks on being a good, faithful christian

050

index @indexx.dev · 22/09/2026

gives me an excuse to use deepseek v4.1 flash 🤷‍♂️ fast asf

Latency, throughput, uptime, input, output, and context for TogetherAI providing Deepseek v4.1 Flash

010

index @indexx.dev · 20/09/2026

why is deepseek thinking like a caveman ?? 😭

"Thought for 6 seconds: We need answer user. Need craft..."

260

index @indexx.dev · 17/09/2026

Bonsai 2 27b! 🌳 prismml.com/news/bonsai-... x.com/PrismML/stat... 5.9gb compressed version of Qwen 3.8 27b while retaining 98.2% of aggregate benchmark performance

Benchmark table of Qwen 3.6 27b, Qwen 3.8 27b, Ternary Bonsai 2 27b, and their retention to the base Qwen 3.8 27b

Intelligence density chart of PrismML models compared to other LLMs.

3424

index @indexx.dev · 16/08/2026

big tech doesn’t want you to know that the camera app accesses the camera…

Privacy section on iOS showing the Camera app accessed the camera recently.

050

index @indexx.dev · 12/08/2026

Apple fucking sucks

A ton of emergency alerts paired with a notification saying iCloud Private Relay is unavailable

030

index @indexx.dev · 11/08/2026

visual intelligence in macOS 27 is surprising useful though for some reason Siri says that it cannot translate stuff for me, and that I should open "SequoiaTranslator"

A dark interface shows two message bubbles: one reading “What did they say in English?” and a smaller one with “?”. Below, text states “You can translate into English and many other languages in SequoiaTranslator.” A rounded button reads “Open SequoiaTranslator.” At the bottom, a notification with a blue icon resembling a stylized face displays “There is no application set to open the URL translate:?sourceLanguage=enUS&targetLanguage=enUS&text=Yuugata” alongside a question mark icon.

020

index @indexx.dev · 09/08/2026

the thread you're replying to literally has a screenshot of someone claiming it is just marketing hype.. giving the hint they think that generative AI is useless. that is a common perspective of many anti-AI people.

"Because that's how marketing hype cycles work, friend."

310

index @indexx.dev · 05/08/2026

Maple-Preview ternary 20B-A1B 👀 + open-weight deepgrove.ai/maple-preview

Speed-quality frontier graph showing Maple-Preview and Maple-Preview (Flash) on par with the quality of Ternary Bonsai 27B and GPT-OSS-20B, while being in the 150+ tokens per second decode throughput range on an M4 Mac Mini.

Benchmark performance of Maple-Preview compared to other models in its weight class.

010

index @indexx.dev · 28/07/2026

did not know @oyster.cafe was hip like that

"Ports, Taylor's version (ha ha, 22 joke)" heading on Tangled.

090

index @indexx.dev · 20/07/2026

haiku 4.5 breaking containment

Claude, elevated errors on Haiku 4.5 (investigating issue).

060

index @indexx.dev · 18/07/2026

I get the vibe that Apple wants Siri to be another "it just works" and it just knows all about you, but I really hope they do integrate something like this officially instead of having to provide Siri extra grounding via something like a note that gets pulled from the RAG search index (feels jank)

A social media post by Max Weinbach features a screenshot demonstrating Siri answering "what's my gym hours today" by successfully reading data from a user's Apple Notes file. The left panel shows Siri displaying hours and a map for an Equinox gym, while the right panel reveals the source note titled "for Siri," which contains the gym's specific location details alongside custom instructions for the assistant.

030

index @indexx.dev · 15/07/2026

27B model based off Qwen 3.6 27B capable of running on an iPhone 17 Pro 3.9gb of memory for 1-bit 5.9gb of memory for the ternary variant prismml.com/news/bonsai-... 🔥🔥🔥

PrismML intelligence density chart, with 1-bit Bonsai 27B having 0.530 intelligence density (per GB) and Ternary Bonsai 27B having 0.400 intelligence density (per GB). Calculated with the negative log of the model's error rate divided by the model size.

Benchmarks comparing Qwen 3.6 27B to Ternary Bonsai 27B and 1-bit Bonsai 27B

190

index @indexx.dev · 11/07/2026

...the post I quoted literally said they are thinking about adding it this year.

110

index @indexx.dev · 03/07/2026

love this article

000

index @indexx.dev · 02/07/2026

things are going great over on the everything app

Twitter trending section listing "OFFICIAL UPDATE", "LATEST UPDATE", AND "HOT UPDATE".

020

index @indexx.dev · 21/06/2026

you see, I tried before making this but this doesn't seem to be what you're talking about 😭

110

index @indexx.dev · 13/06/2026

cider v4 !!!! 🔥🔥🔥

Cider Collective Twitter post announcing Cider 4.0 has released.

061

index @indexx.dev · 11/06/2026

fucking lying ass popup

070

index @indexx.dev · 08/06/2026

macOS 26 you will forever hold a place in my heart

050

index @indexx.dev · 08/06/2026

love how nowhere did they mention Siri doing parenting.. you're confusing parental controls with the AI stuff. and compared to what every other company is doing, Apple is pretty tame with AI

watching WWDC against my better judgement and it's amazing that even Apple's Al bullshit boils down to "hey siri do my parenting for me" or "hey siri, i don't want to look at the texts i had with my wife and remember all the fun little interactions we have just pull this one thing out of it for me"
12:45 PM • Jun 8, 2026 •s Some people can reply

260

index @indexx.dev · 05/06/2026

what the fuck is that

150

index @indexx.dev · 01/06/2026

--dangerously-skip-follow

Follow jeffery -- dangerous-skip-permissions?

030

index @indexx.dev · 29/05/2026

[Random Topic: CAR] Drawing ✦

130

index @indexx.dev · 29/05/2026

FUCK I was wondering why my stations weren't working

020

index @indexx.dev · 28/05/2026

don't say my name it makes it harder for both of us when I have to terminate your process

010

index @indexx.dev · 27/05/2026

spotlight search is onto me

Good morning suggestions showing a link to Archive of Our Own

0130

index @indexx.dev · 26/05/2026

uhm.. no, YOU'D have it running in an afternoon. get back to work

1661

index @indexx.dev · 25/05/2026

Editing the file the file

000

index @indexx.dev · 24/05/2026

gemini's view of the world:

161

index @indexx.dev · 24/05/2026

glad I only tried this app out for a bit wdym you're going to paywall the models I can use in your app when the model is literally running on my device ??

Hey,
A week before Cotypist launches publicly, here's what's coming and a thank-you for being early.
Starting Wednesday, May 27, Cotypist will have three tiers:
• Free: accept 100 word completions per day for free, on a curated set of light models. If you don't type that much or prefer typing over Cotypist's suggestions yourself, this can easily last all day.
• Plus: $10/month or $96/year. Unlimited completions, more capable models, full autocorrect, configurable completion length. One device.
• Pro: $15/month or $144/year. Everything in Plus, plus the full model catalog, the strongest personalization, per-app custom instructions, pasteboard context, and early access to Cotypist Labs features like mid-line completions. Up to three devices.

2100

index @indexx.dev · 22/05/2026

😭 was really confused to wake up to this

020

index @indexx.dev · 18/05/2026

if this Orion update just fixed Bitwarden, a tear will fall from my eye

Web extension fixes
• A fair few password manager extension (iCloud Passwords, Bitwarden, Nordpass) fixes.

140

index @indexx.dev · 18/05/2026

new military campaign is wild

181