๐•ญ๐–—๐–Ž๐–†๐–“ ๐•ฎ๐•ฒ (@_briancg) 's Twitter Profile
๐•ญ๐–—๐–Ž๐–†๐–“ ๐•ฎ๐•ฒ

@_briancg

โš”๏ธ DS | ML Engineer
๐Ÿš€ Cofounder - parseai.co

ID: 2501362603

linkhttps://bit.ly/3dYxNcO calendar_today17-05-2014 11:40:35

2,2K Tweet

230 Followers

959 Following

CyberSudo (@cyber_sudo) 's Twitter Profile Photo

See How the World Has Changed Over Time ๐ŸŒ๐Ÿ›ฐ๏ธ Ever wondered how cities, infrastructure, or landscapes evolved over the years? World Imagery Wayback makes it easy. This free service archives satellite images since 2014, allowing you to track territorial changes over time by

elvis (@omarsar0) 's Twitter Profile Photo

Too many people working with multi-agent systems assume that if you just add enough agents and let them talk, interesting social dynamics will emerge. A new paper suggests that the assumption is fundamentally wrong. Researchers studied Moltbook, a social network with no humans,

Too many people working with multi-agent systems assume that if you just add enough agents and let them talk, interesting social dynamics will emerge.

A new paper suggests that the assumption is fundamentally wrong.

Researchers studied Moltbook, a social network with no humans,
๐•ญ๐–—๐–Ž๐–†๐–“ ๐•ฎ๐•ฒ (@_briancg) 's Twitter Profile Photo

When writing, we optimize our learning rates. When using LLMs to write, we optimize theirs at the expense of our own. For years, I've been writing for myself: findings, mistakes, wins, losses, breakthroughs. Never shared outside work, though. Maybe I should, we'll see.

Unsloth AI (@unslothai) 's Twitter Profile Photo

100,000+ models trained with Unsloth have now been open-sourced on Hugging Face! Popular fine-tuned LLMs you can run locally: 1. TeichAI - GLM-4.7-Flash distilled from Claude 4.5 Opus (high) 2. Zed - Qwen Coder 7B fine-tuned for stronger coding 3. DavidAU - Llama-3.3-8B

100,000+ models trained with Unsloth have now been open-sourced on Hugging Face!

Popular fine-tuned LLMs you can run locally:
1. TeichAI - GLM-4.7-Flash distilled from Claude 4.5 Opus (high)
2. Zed - Qwen Coder 7B fine-tuned for stronger coding
3. DavidAU - Llama-3.3-8B
Xinghong (Shin) Fu (@shinfxh) 's Twitter Profile Photo

Do we really need billion-parameter transformers for time series forecasting? What if we did it in under 3M parameters? Introducing Reverso: time series foundation models as small as 200K paramters significantly pushing the Pareto frontier in zero-shot time series forecasting!

Do we really need billion-parameter transformers for time series forecasting?

What if we did it in under 3M parameters?

Introducing Reverso: time series foundation models as small as 200K paramters significantly pushing the Pareto frontier in zero-shot time series forecasting!
Shaded Maps (@shadedmaps) 's Twitter Profile Photo

Diagonal panning video from a very detailed shaded map of Genova, Italy. The full map is 36000 x 16700 pixels with 0.5 m resolution and is available at shadedmaps.github.io Data source: mappe.comune.genova.it [Comune di Genova]. #maps #shadedmaps #cartography #shadedrelief

elvis (@omarsar0) 's Twitter Profile Photo

New Google paper challenges how we measure LLM reasoning. Token count is a poor proxy for actual reasoning quality. There might be a better way to measure this. This work introduces "deep-thinking tokens," a metric that identifies tokens where internal model predictions shift

New Google paper challenges how we measure LLM reasoning.

Token count is a poor proxy for actual reasoning quality.

There might be a better way to measure this.

This work introduces "deep-thinking tokens," a metric that identifies tokens where internal model predictions shift
Tailscale (@tailscale) 's Twitter Profile Photo

Use local models on remote devices you controlโ€”as if they were local. - Introducing LM Link from @LMStudio. - Encrypted, identity-based access to your own LLM hardware. - No public endpoints. No API key sprawl. tailscale.com/blog/lm-link-rโ€ฆ

Use local models on remote devices you controlโ€”as if they were local.

- Introducing LM Link from @LMStudio.
- Encrypted, identity-based access to your own LLM hardware.
- No public endpoints. No API key sprawl.

tailscale.com/blog/lm-link-rโ€ฆ
Muhammad Ayan (@socialwithaayan) 's Twitter Profile Photo

๐Ÿšจ BREAKING: Alibaba just handed the AI agent community a production-grade sandbox for free. OpenSandbox is a full-stack platform for running untrusted agent code safely: โ†’ Unified APIs across multi-language SDKs โ†’ Docker and Kubernetes runtimes purpose-built for agents โ†’

๐Ÿšจ BREAKING: Alibaba just handed the AI agent community a production-grade sandbox for free.

OpenSandbox is a full-stack platform for running untrusted agent code safely:

โ†’ Unified APIs across multi-language SDKs
โ†’ Docker and Kubernetes runtimes purpose-built for agents
โ†’
Prof. Anima Anandkumar (@animaanandkumar) 's Twitter Profile Photo

Weโ€™re excited to release TorchLean which is the first fully verified neural network framework in Lean. The Lean community has largely focused on pure mathematics. TorchLean expands this frontier toward verified neural network software and scientific computing. With the recent

Weโ€™re excited to release TorchLean which is the first fully verified neural network framework in Lean. The Lean community has largely focused on pure mathematics. TorchLean expands this frontier toward verified neural network software and scientific computing. With the recent
Sergio Paniego (@sergiopaniego) 's Twitter Profile Photo

Qwen3.5 dense (smol ๐Ÿค) models just dropped - natively multimodal - 0.8B ยท 2B ยท 4B ยท 9B (+ base variants) - 262K context extensible to 1M - built-in thinking fine-tune them with TRL out of the box โ†’ SFT, GRPO, DPO and more!

Qwen3.5 dense (smol ๐Ÿค) models just dropped

- natively multimodal
- 0.8B ยท 2B ยท 4B ยท 9B (+ base variants)
- 262K context extensible to 1M
- built-in thinking

fine-tune them with TRL out of the box โ†’ SFT, GRPO, DPO and more!
Junyi Zhang (@junyi42) 's Twitter Profile Photo

๐—ข๐—ป๐—ฒ ๐—บ๐—ฒ๐—บ๐—ผ๐—ฟ๐˜† ๐—ฐ๐—ฎ๐—ปโ€™๐˜ ๐—ฟ๐˜‚๐—น๐—ฒ ๐˜๐—ต๐—ฒ๐—บ ๐—ฎ๐—น๐—น. We present ๐—Ÿ๐—ผ๐—š๐—ฒ๐—ฅ, a new ๐—ต๐˜†๐—ฏ๐—ฟ๐—ถ๐—ฑ ๐—บ๐—ฒ๐—บ๐—ผ๐—ฟ๐˜† architecture for long-context geometric reconstruction. LoGeR enables stable reconstruction over up to ๐Ÿญ๐Ÿฌ๐—ธ ๐—ณ๐—ฟ๐—ฎ๐—บ๐—ฒ๐˜€ / ๐—ธ๐—ถ๐—น๐—ผ๐—บ๐—ฒ๐˜๐—ฒ๐—ฟ ๐˜€๐—ฐ๐—ฎ๐—น๐—ฒ, with

Andrej Karpathy (@karpathy) 's Twitter Profile Photo

Software horror: litellm PyPI supply chain attack. Simple `pip install litellm` was enough to exfiltrate SSH keys, AWS/GCP/Azure creds, Kubernetes configs, git credentials, env vars (all your API keys), shell history, crypto wallets, SSL private keys, CI/CD secrets, database