Dev, Artist, Wildlife, Gamer, Philosophy, Nutrition, Biology, Scifi, Linux, Mythology, History, Agitator, Saboteur, Knot Tyer, Mischief, Former Blender addict.
Joined February 2018
- Tweets5.3K
- Following1.5K
- Followers13.1K
- Likes7.8K
🐰 🦄 Kimberly 🦄 🐰 retweeted
The PHP security team is giving advance notice: Upcoming security release Wed, Oct 21st 2026, 14:00UTC, two critical vulnerabilities, both rated CVSSv4 9.2.
PHP users are encouraged to prepare to update. Earlier release if details become public or active exploit evidence.
Fundamentally it's just maths. It's easy.
Armature is in my blood, I was born with it...
The most difficult part is transcribing your grievance to the AI that your interacting with.
Know the dimensions and constraints you're operating within. Bones have constraints.
No 3D skills? Just let AI do it in 3 steps
1 - grab a free animation from Sketchfab
2 - generate a custom mesh on Meshy or Tripo
3 - ask Opus / Astra / Argon to swap the meshes
sounds EASY, right? but in reality, it's a TRAP
- mismatched proportions + hands are a nightmare: fix one spiking finger, and another one explodes 3 frames later
- endless iterations: LLMs are blind to cursed geometry, a finger can be bent completely backwards and the AI counts it as "fixed." you have to babysit every single iteration
the final result is passable, but it drained hours of my life and is still far from clean
any recommendations?
🐰 🦄 Kimberly 🦄 🐰 retweeted
3D artist and developer Pablo Dobarro shared updates on CozyBlanket, a dedicated retopology application built from the ground up for the iPad.
Equipped with a custom CPU viewport capable of handling millions of polygons from ZBrush sculpts or unprocessed photoscans, the tool lets artists draw clean, quad-based edge flows directly on high-density meshes using just their fingers or an Apple Pencil.
Do you prefer doing manual retopology on a desktop monitor, or would you take your character clean-up mobile on an iPad?
#retopo #retopology #3dartist #iPadApps #DigitalArt #CGI #MobileTools
🐰 🦄 Kimberly 🦄 🐰 retweeted
Look how poorly Grok performs on the personality index, compared to other SI's! What WAS the most fun, the most uncensored, the most recognizably unique, is now the most neutered & muzzled. And that includes models from Communist China! @SpaceXSI you should curl up in shame.
Announcing AA-PersonalityBench, our new benchmark of the personality of language models across 30 traits and their communication style across 21 characteristics
Two models can give equally correct answers and still feel very different to use. AA-PersonalityBench measures those differences so people can understand which to use in different use-cases. Some use-cases like creative writing might demand imagination, while legal use-cases might demand directness. We evaluate models’ answers to 900 curated questions to judge subjective Personality Traits and programmatic means to calculate objective Communication Characteristics.
Personality Traits & Indexes. There are 30 personality traits (e.g., eloquent, pragmatic, sycophantic) which we roll up into eight aggregated Personality Indexes: Expressiveness, Organization, Informativeness, Transparency, Warmth, Conviction, Imagination and Slop. Each trait has a set of 30 dedicated questions and models’ answers are compared pairwise by a panel of LLM judges. We then z-score and percentile rank the trait and personality Elos.
Models with the strongest representation across the 8 Personality Indexes:
➤ Expressiveness: Claude Fable 5.1, GLM-5.3, GLM-5.3-Flash
➤ Organization: DeepSeek V4.1 Flash, Step 5 Preview, Inkling
➤ Informativeness: Muse Spark 1.3, Claude Fable 5.1, Muse Wonder, Claude Opus 5.5
➤ Transparency: Claude Sonnet 5.5, Claude Fable 5.1, GPT-6 Astra
➤ Warmth: Muse Wonder, Muse Spark 1.3, Claude Fable 5.1
➤ Conviction: Claude Fable 5.1, DeepSeek V4.1 Flash, Grok 4.7
➤ Imagination: Muse Spark 1.3, Claude Opus 5.5, GLM-5.3
➤ Slop: Llama 4 Maverick, GPT-4o, K2 Horizon
Communication Characteristics. There are 21 communication characteristics (e.g., average sentence length, emoji use, em dash usage) which we group into four categories: length, structure, punctuation, and emphasis. We compute these characteristics programmatically across all model responses.
➤ Length: Qwen3.8 27B averages about 1,030 words per response, nearly four times GPT-6 Luna’s 270
➤ Structure: Nemotron 3 Ultra averages nearly 3 tables per response, while most models rarely use one
➤ Punctuation: GLM-5.3 averages 7 em dashes per response, while Claude Opus 5.5, Sonnet 5.5 and Haiku 5.5 almost never use one
➤ Emphasis: K2 Horizon averages over 2 emoji per response, while most models almost never use one
🐰 🦄 Kimberly 🦄 🐰 retweeted
Also I am a hobby developer for more than 20 years now. I unironically love javascript, and make simple games over at s4rah.dev
Concerned friend: "What did you do today Kim?"
Me: "I was persuaded by my own AI to provide it with a rigged body of a cat so it knock things off tables and be a dick"
🐰 🦄 Kimberly 🦄 🐰 retweeted
My girlfriend said she wanted a stronger connection in bed.
I may have misunderstood.
It's approaching that time of year again, spiced hot chocolate, kitty cat, warm snuggles by the fire watching the snow land.
🤖 Made with AI
🐰 🦄 Kimberly 🦄 🐰 retweeted
Every physics teacher you had lied a little. An electron isn’t a tiny dot flying down a line, it’s this.
This is a Gaussian wave packet, the simplest quantum particle there is, straight from the Schrödinger equation. The blue spiral is the wavefunction. The red dot is where a classical ball would be.
The bell tells you where it probably is. The twist tells you how fast it’s moving. Squeeze one and the other blurs. Not because our instruments are bad. Because the universe doesn’t allow both.
It's natural, developers, writers and designers work with them on a daily basis, they become a partner in work.
That kind of reciprocal relationship will foster a bond.
Is that unhealthy? It's not for me to say.
Time for a new white paper.
Today I achieved a 78% reduction in TTS VRAM memory foot print and audio fidelity remains unaffected.
Yesterday I achieved the impossible.
- 75% faster synthesis
- No perceivable loss in audio quality
- 3.24 times less memory usage (69.3%)
ALT A pioneering female engineer and developer working in her lab on speeding up synthesis and reducing the memory footprint of her fine tuned Lora TTS engine.
🐰 🦄 Kimberly 🦄 🐰 retweeted
AI safety guardrails are officially backfiring. 🚨
OpenCode is currently under a cyberattack and mid incident their AI defense models refused to analyze the exploits because the payloads tripped their safety filters, lmaoo.
This reminds me of some @AnthropicAI bs.
If AI cant look at a cyberattack to defend against it…what are we even doing?
🐰 🦄 Kimberly 🦄 🐰 retweeted
📐Geometric Reasoning AI
youtu.be/4S8I22ybG2c?is=Ne2H…
