Kol Tregaskes
@koltregaskes.bsky.social
AI Art Creator 🎨 | AI Video Producer 🎬🤖 | AI Music Composer 🎵 | Tech & Science News Curator🔬 | Movie & Sci-fi Enthusiast 🍿 | Geek at Heart 🤓 | #AI #Tech
OpenAI counters Musk's lawsuit with 2017-2018 evidence showing he advocated for-profit transition.
- Musk agreed to a hybrid philanthropic for-profit model, similar to OpenAI's current public benefit corporation structure valued at $130B.
- Musk agreed to a hybrid philanthropic for-profit model, similar to OpenAI's current public benefit corporation structure valued at $130B.
January 17, 2026 at 3:31 AM
OpenAI counters Musk's lawsuit with 2017-2018 evidence showing he advocated for-profit transition.
- Musk agreed to a hybrid philanthropic for-profit model, similar to OpenAI's current public benefit corporation structure valued at $130B.
- Musk agreed to a hybrid philanthropic for-profit model, similar to OpenAI's current public benefit corporation structure valued at $130B.
The ads are coming.
Only in free and Go - logged in over 18 - plans, only the US for now.
x.com/i/status/20...
Only in free and Go - logged in over 18 - plans, only the US for now.
x.com/i/status/20...
January 17, 2026 at 12:29 AM
The ads are coming.
Only in free and Go - logged in over 18 - plans, only the US for now.
x.com/i/status/20...
Only in free and Go - logged in over 18 - plans, only the US for now.
x.com/i/status/20...
Results from AA-Omniscience benchmark show no single best AI model for knowledge across programming languages.
- Claude 4.5 Opus leads in Python with 56 points and Go with 54.
- Gemini 3 Pro Preview tops JavaScript with 56 and Swift with 56.
- Claude 4.5 Opus leads in Python with 56 points and Go with 54.
- Gemini 3 Pro Preview tops JavaScript with 56 and Swift with 56.
January 16, 2026 at 9:31 PM
Results from AA-Omniscience benchmark show no single best AI model for knowledge across programming languages.
- Claude 4.5 Opus leads in Python with 56 points and Go with 54.
- Gemini 3 Pro Preview tops JavaScript with 56 and Swift with 56.
- Claude 4.5 Opus leads in Python with 56 points and Go with 54.
- Gemini 3 Pro Preview tops JavaScript with 56 and Swift with 56.
Claude AI achieves 9-12x productivity gains on tasks needing 14.4 years education, per Anthropic's Economic Index from November 2025 data.
- Primitives measure complexity, skills (14.4 years average vs. economy's 13.2), use cases (46% work-related), autonomy (3.5/5), and success (67% on Claude).
- Primitives measure complexity, skills (14.4 years average vs. economy's 13.2), use cases (46% work-related), autonomy (3.5/5), and success (67% on Claude).
January 16, 2026 at 6:30 PM
Claude AI achieves 9-12x productivity gains on tasks needing 14.4 years education, per Anthropic's Economic Index from November 2025 data.
- Primitives measure complexity, skills (14.4 years average vs. economy's 13.2), use cases (46% work-related), autonomy (3.5/5), and success (67% on Claude).
- Primitives measure complexity, skills (14.4 years average vs. economy's 13.2), use cases (46% work-related), autonomy (3.5/5), and success (67% on Claude).
Great to see Anthropic's skills standard is spreading. Now in Google's Antigravity IDE.
x.com/antigravity...
x.com/antigravity...
January 16, 2026 at 3:33 PM
Great to see Anthropic's skills standard is spreading. Now in Google's Antigravity IDE.
x.com/antigravity...
x.com/antigravity...
January 16, 2026 at 3:07 PM
ChatGPT has improved chat history
x.com/_samirism/s...
x.com/_samirism/s...
January 16, 2026 at 9:29 AM
ChatGPT has improved chat history
x.com/_samirism/s...
x.com/_samirism/s...
Elon says Grok 4.20 *won't* beat Opus 4.5 in coding but will be good in other areas.
x.com/i/status/20...
x.com/i/status/20...
January 16, 2026 at 6:49 AM
Elon says Grok 4.20 *won't* beat Opus 4.5 in coding but will be good in other areas.
x.com/i/status/20...
x.com/i/status/20...
GLM-4.7 from Z AI outperforms DeepSeek V3.2 in agentic benchmarks like Tau2 Telecom, released as an open-weight flagship in December 2025, with high token consumption elevating speed and cost factors.
January 16, 2026 at 12:34 AM
GLM-4.7 from Z AI outperforms DeepSeek V3.2 in agentic benchmarks like Tau2 Telecom, released as an open-weight flagship in December 2025, with high token consumption elevating speed and cost factors.
Oh man, when did Anthropic release all these in Claude? Control Chrome, Windows MCP and Filesystem, let's go!
January 15, 2026 at 9:34 AM
Oh man, when did Anthropic release all these in Claude? Control Chrome, Windows MCP and Filesystem, let's go!
Google launches Personal Intelligence beta in the US, integrating Gmail, Photos, YouTube, and Search data for tailored Gemini app responses like trip planning or shopping suggestions.
January 15, 2026 at 12:26 AM
Google launches Personal Intelligence beta in the US, integrating Gmail, Photos, YouTube, and Search data for tailored Gemini app responses like trip planning or shopping suggestions.
GLM-Image from z ai beats Nano Banana Pro and GPT-image 1.5 on text rendering benchmarks.
- Released as open-source with hybrid autoregressive and diffusion decoder architecture.
- Released as open-source with hybrid autoregressive and diffusion decoder architecture.
January 14, 2026 at 9:33 PM
GLM-Image from z ai beats Nano Banana Pro and GPT-image 1.5 on text rendering benchmarks.
- Released as open-source with hybrid autoregressive and diffusion decoder architecture.
- Released as open-source with hybrid autoregressive and diffusion decoder architecture.
This is very important, adjust the retention length or I got Claude Code to export the conversations itself regularly.
x.com/i/status/20...
x.com/i/status/20...
January 14, 2026 at 6:33 PM
This is very important, adjust the retention length or I got Claude Code to export the conversations itself regularly.
x.com/i/status/20...
x.com/i/status/20...
1X NEO releases 1XWM world model, pretrained on web videos, tuned on 900 hours human data and 70 hours robot data, achieving 45% success on novel tasks via parallel video generations.
January 14, 2026 at 3:58 PM
1X NEO releases 1XWM world model, pretrained on web videos, tuned on 900 hours human data and 70 hours robot data, achieving 45% success on novel tasks via parallel video generations.
Meta announces Meta Compute initiative to build tens of gigawatts of AI infrastructure this decade.
- Meta is establishing a new top-level initiative called Meta Compute.
- How they engineer, invest, and partner to build this infrastructure will become a strategic advantage.
- Meta is establishing a new top-level initiative called Meta Compute.
- How they engineer, invest, and partner to build this infrastructure will become a strategic advantage.
January 14, 2026 at 3:31 PM
Meta announces Meta Compute initiative to build tens of gigawatts of AI infrastructure this decade.
- Meta is establishing a new top-level initiative called Meta Compute.
- How they engineer, invest, and partner to build this infrastructure will become a strategic advantage.
- Meta is establishing a new top-level initiative called Meta Compute.
- How they engineer, invest, and partner to build this infrastructure will become a strategic advantage.
Grok 4.2 models on LM Arena? Release can't be far off. CLI also rumoured.
slateflow and tidewisp
x.com/i/status/20...
slateflow and tidewisp
x.com/i/status/20...
January 14, 2026 at 3:15 PM
Grok 4.2 models on LM Arena? Release can't be far off. CLI also rumoured.
slateflow and tidewisp
x.com/i/status/20...
slateflow and tidewisp
x.com/i/status/20...
I keep wanting to say "Claude CoworkER!" :-)
January 14, 2026 at 12:26 PM
I keep wanting to say "Claude CoworkER!" :-)
My agents built this in Obsidian as the start of my hub. Long-term, I hope to eliminate the need for buttons. We're also developing a second brain system. I was deep into my version when I saw Nate B. Jones' video and am now trying to incorporate some of his ideas.
January 14, 2026 at 9:27 AM
My agents built this in Obsidian as the start of my hub. Long-term, I hope to eliminate the need for buttons. We're also developing a second brain system. I was deep into my version when I saw Nate B. Jones' video and am now trying to incorporate some of his ideas.
OpenAI's Sweetpea earbuds leak details Jony Ive-designed AI device to replace AirPods.
- Special audio product codenamed Sweetpea, prioritised by Jony Ive's team.
- Design features metal eggstone shape housing two removable pill capsules that rest behind the ear.
- Special audio product codenamed Sweetpea, prioritised by Jony Ive's team.
- Design features metal eggstone shape housing two removable pill capsules that rest behind the ear.
January 14, 2026 at 7:29 AM
OpenAI's Sweetpea earbuds leak details Jony Ive-designed AI device to replace AirPods.
- Special audio product codenamed Sweetpea, prioritised by Jony Ive's team.
- Design features metal eggstone shape housing two removable pill capsules that rest behind the ear.
- Special audio product codenamed Sweetpea, prioritised by Jony Ive's team.
- Design features metal eggstone shape housing two removable pill capsules that rest behind the ear.
Oh no! Maybe it's time for Max x 20?
January 13, 2026 at 11:58 PM
Oh no! Maybe it's time for Max x 20?
Is Claude Cowork just Claude Code Desktop with a cleaner UI? This is great, but as a Windows user who already uses Claude Desktop, am I missing out on anything yet?
January 13, 2026 at 9:27 PM
Is Claude Cowork just Claude Code Desktop with a cleaner UI? This is great, but as a Windows user who already uses Claude Desktop, am I missing out on anything yet?
Claude Cowork was built in just a a week and a half,.**1.5 WEEKS!** And all the code was written by their coding agent, Claude Code.
This is the world we live in right now. 🤯
This is the world we live in right now. 🤯
January 13, 2026 at 6:31 PM
Claude Cowork was built in just a a week and a half,.**1.5 WEEKS!** And all the code was written by their coding agent, Claude Code.
This is the world we live in right now. 🤯
This is the world we live in right now. 🤯
PixVerse R1, a realtime world model, released, generates high-resolution video in real-time with infinite streaming.
- It unifies text, image, video, and audio into a single token stream for seamless multimodal processing.
- It unifies text, image, video, and audio into a single token stream for seamless multimodal processing.
January 13, 2026 at 3:48 PM
PixVerse R1, a realtime world model, released, generates high-resolution video in real-time with infinite streaming.
- It unifies text, image, video, and audio into a single token stream for seamless multimodal processing.
- It unifies text, image, video, and audio into a single token stream for seamless multimodal processing.
Amazingly, Anthropic beat OpenAI to this!
x.com/claudeai/st...
x.com/claudeai/st...
January 13, 2026 at 4:59 AM
Amazingly, Anthropic beat OpenAI to this!
x.com/claudeai/st...
x.com/claudeai/st...
January 13, 2026 at 3:31 AM