Audio, generative AI & Python workflows. Building Logic Pro MCP, FL Studio MCP and Claude Animation Studio. Practical tools for creative work.
Joined December 2025
- Tweets217
- Following33
- Followers18
- Likes1
try this if you mix in fl studio. i built a personal sound engineer for it: an open-source mcp that works alongside you in your project.
copy this into claude once it's connected:
"scan my mix and tell me what's wrong. show the evidence for each problem and wait for my ok before changing anything."
mix doctor reports clipping, low headroom, level imbalance, missing high-pass and overlapping eq boosts. you stay the producer: every fix is shown first, applied one at a time, and can be rolled back.
it also writes in ragas and maqam, not just western scales.
the caveat: on aa-omniscience it answered 56% of questions correctly, but when it didn't know, it almost always guessed instead of saying so. that's a 93% hallucination rate under aa's definition. this does not mean 93% of all its answers were wrong.
if you use it in an agent, add retrieval and checks. minimax says open weights are coming oct 23.
sources:
minimax blog: minimax.io/blog/minimax-m31f
venturebeat (artificial analysis results, weights date): venturebeat.com/technology/m…
minimax announcement: xitter.cf/MiniMax_AI/status/2112…
Thanks for keeping the secret, @OpenRouter 🤫🐰
And thank you to every dev who kept Space Bunny at #1 for 10 straight days 🏆
You were using the early build. Wait till you try this one: The final M3.1 Flash is live and free on OpenRouter for 72h!
india ai daily 🇮🇳 last night's launches, sorted by "can i actually use this here today?"
→ claude duet: not yet. us pro & max waitlist only
→ claude security auto re-scans: only on claude enterprise (public beta)
→ recursive flint, a 40b coding model: yes. free weights, apache-2.0, unsloth says it runs on 24gb ram
→ openai's php2shell checker: yes. free repo, checks your php version
the one to try today is flint. ollama got it running with one line: ollama run flint
sources:
claude duet (us pro/max waitlist): support.claude.com/en/articl…
claude security: xitter.cf/claudeai/status/211296…
flint: recursive.com/updates/introd…
flint 24gb (unsloth): xitter.cf/UnslothAI/status/21130…
php2shell checker: github.com/openai/php2shell-…
Run Flint locally on 24GB RAM via Unsloth Dynamic GGUFs!
Flint is the strongest MoE model under 50B parameters.
Recursive’s Apache-2.0 model has 39.8B parameters (2.5B active). Run at 125 t/s on Mac, 150 on RTX Spark
GGUF: huggingface.co/unsloth/flint…
Guide: unsloth.ai/docs/models/flint
india ai daily 🇮🇳 last night's launches, sorted by "can i actually use this here today?"
→ claude duet: not yet. us pro & max waitlist only
→ claude security auto re-scans: only on claude enterprise (public beta)
→ recursive flint, a 40b coding model: yes. free weights, apache-2.0, unsloth says it runs on 24gb ram
→ openai's php2shell checker: yes. free repo, checks your php version
the one to try today is flint. ollama got it running with one line: ollama run flint
the voice ai price war just got real. here's what launched this week, side by side:
→ elevenlabs realtime v1 max: 0.57s to first audio
→ elevenlabs realtime v1 mini: 0.62s to first audio
→ both cost $0.08/min of connected audio plus tokens (mini $0.20 in / $6 out per 1m, max $1.80 / $9)
→ mistral voxtral 3 voice agent api: $0.04/min with mistral small 4, $0.08/min with a larger model. mistral says each turn finishes in under a second
→ india data residency: only v1 mini is available there, not v1 max
for teams in india the choice is simple: if your data must stay in india, it's mini or something else. latency and speed figures are from the vendors.
sources:
eleven realtime launch: elevenlabs.io/blog/realtime
pricing: elevenlabs.io/pricing/api
india/singapore residency: elevenlabs.io/docs/eleven-ag…
voxtral 3 + pricing: mistral.ai/news/voxtral-3/
gtc berlin, the 3 launches builders should care about:
→ openshell reference architecture: agents run on vera rubin, harnesses on vera cpus, bluefield-4 + nvidia sentry stops an agent that leaves its boundary in milliseconds. available now
→ dsx pod: vera rubin ai factories built in containers, deployed at a prepared site in as little as 14 days vs 6 months. ships early 2027
→ isaac lab 3.0 is GA: open source robot learning, desktop to data center
agent guardrails in silicon: real need or vendor lock-in?
sources:
openshell ra: nvidianews.nvidia.com/news/o…
dsx pod: nvidianews.nvidia.com/news/n…
isaac: nvidianews.nvidia.com/news/n…
keynote: youtube.com/live/PvFjbPYEOZI
openai's models are now coming to all 3 big clouds. google cloud is the last one.
→ openai frontier models (gpt-6 astra) + codex land in google cloud's model garden
→ buy through google cloud marketplace and draw down your existing gcp commitments
→ also coming soon to google antigravity and the gemini agent
→ aws got openai models in april, right after microsoft's exclusivity ended
the catch: nothing is live yet. GA is "planned in the coming weeks", and microsoft stays openai's primary cloud partner
does this make you pick a model, or just a cloud?
openai.com/index/openai-mode… #rosasynthesiz
sources:
openai: openai.com/index/openai-mode…
reuters via cna: channelnewsasia.com/business…
cnbc: cnbc.com/2026/10/21/openai-b…
the company behind PUBG and BGMI just released its first AI model. and you can download it.
raon-1 mini (preview) from krafton:
→ 30B MoE, only 4.3B active per token, so it's fast
→ apache 2.0 open weights, commercial use allowed
→ 262K context, built-in thinking, tool calling
→ runs on one RTX 5090 (32GB) in NVFP4. GGUF and Mac MLX builds too
→ krafton's own chart: 70.8 on SWE-bench Verified vs 72.0 for Muse Glimmer 30B
a game studio training a model from scratch is the real story. would you put it in your stack?
sources: xitter.cf/Krafton_AI/status/2112… · huggingface.co/KRAFTON/Raon-… (preview release, specs may change)
Today, we’re releasing a preview of Raon-1 mini, a 30B MoE LLM with 4B active parameters, trained from scratch by KRAFTON AI.
With our efficient architecture, it matches Muse Glimmer on coding/agentic tasks while being much faster and more memory-efficient! 🔥 Plus, natural, human-like conversation.
This is just a first look. The final release (and bigger models) are coming soon! 👀
👉 Try it now, free for a limited time: raon-mini.krafton.ai/en
🤗 Open weights + DFlash drafter (Apache 2.0): huggingface.co/KRAFTON/Raon-…
