Community Updates
A stream of short-form insights, takes, and news from the AI Socratic community.
GradMem: Writing Context into LLM Memory via Test-Time Gradient Descent
GradMem: Writing Context into LLM Memory via Test-Time Gradient Descenthttps://x.com/yurakuratov/status/2034989528818098615, GradMem introduces writing context into memory using test-time gradient des
100M Token Context Without Collapse on 2×A800 GPUs
100M Token Context Without Collapse: <9% Degradation on 2×A800 GPUshttps://x.com/marmaduke091/status/2034736279884025916, New research achieves 100M token context windows with less than 9% degradation
LLM Internals: By Layer 10, Models Are Language-Agnostic
LLM Internals: By Layer 10, Models Don't Know What Language They're Readinghttps://x.com/dnhkng/status/2036187044704035089, A new blog post reveals that when feeding the same sentence in English and C
LLM Fused with Mini Computer: Switching Between Text and Machine Code
LLM Fused with Mini Computer: Switching Between Text and Machine Code in Single GPUhttps://x.com/EastlondonDev/status/2036759710070645071, A developer demonstrates an LLM brain fused with a mini compu
Columbia Exposes Flaws in Private AI Inference: 280GB per Query
Columbia University Exposes Flaws in Private AI Inference: Prior Methods Used 280GB per Queryhttps://x.com/godofprompt/status/2035697480227266950, Columbia University researchers prove that the entire
LiteLLM PyPI Supply Chain Attack Exfiltrates Credentials
LiteLLM's PyPI release 1.82.8 was compromised in a major supply chain attack. A simple pip install litellm could exfiltrate SSH keys, AWS/GCP/Azure credentials, Kubernetes configs, API keys, crypto wa
ARC-AGI-3 Announced: Humans Score 100%, AI < 1%
This is so far the only unsaturated agentic intelligence benchmark. Unlike benchmarks that test what models already know, ARC-AGI-3 tests how they learn and acquire new skills, providing a formal meas
Apple Opening Up Siri to Other Models
Apple Opening Up Siri To Other tweet id="2037230804942610548"
Quantization Explained
Quantization Explained tweet id="2036844409145512255"
Gemini Embedding 2: Natively Multimodal Embedding Model
Gemini Embedding 2 Gemini Embedding 2 is our first natively multimodal embedding model that maps text, images, video, audio and documents into a single embedding space, enabling multimodal retrieval a
The State of AI Safety in 4 Fake Graphs
The State of AI Safety in 4 Fake Graphs tweet id="2038606572046172443"
China Has 339 GW of Wind and Solar Under Construction
China currently has 339 gigawatts of wind and solar capacity under construction
OpenAI GPT-5.4 (xhigh) Released
Offers roughly the same benchmark performance as Gemini 3.1 Pro, but for ~25% $USD/M tokens. Sources: Artificial Analysishttps://artificialanalysis.ai/models/gpt-5-4
Google Gemini 3.1 Flash-Lite
This is the fastest lightweight model. Google has been releasing the Flash model shortly after releasing the Pro models, Jeff Dean in the Latent Space Pod confirmed that the flash models are a distill
Alibaba Qwen 3.5 Small Model Series
Introducing Qwen 3.5 Small Model Series: Qwen3.5-0.8B · Qwen3.5-2B · Qwen3.5-4B · Qwen3.5-9B. These small models are built on the same Qwen3.5 foundation — native multimodal, improved architecture, sc
xAI Grok 4.20 with Parallel Agents
xAI new version of Grok runs 4 Grok4 agents in parallel. The result is not too bad. xAI added a new SuperGrok Heavy tier that runs 16 agents. While Grok is still far from OpenAI and Anthropic level, i
StepFun's Step 3.5 Flash
Sparse MoE model with 196B total params, but only 11B activated per token, this model was designed to fit into 128 GB memory i.e. it can run on DGX spark or other local setups. It is one of the first
Anthropic Draws a Red Line with the Pentagon
One of the most consequential AI policy fights of the year erupted between Anthropic and the U.S. Department of War. The standoff triggered a broader divide across the AI industry. Anthropic framed th
OpenAI, xAI, and Google Sign DoW Defense Agreements
Sama first seconded Dario, the next day signed a DoW agreement, pledging similar safeguards but agreeing to provide models for defense use under a $120M contract. That decision wasn't well received by
AWS UAE Data Center Bombed
In the early hours of March 1, 2026, amid Iran's retaliatory drone and missile strikes across the Gulf following US and Israeli attacks on Tehran, Amazon Web Services' ME-CENTRAL-1 region in the UAE t
What happens to AI when Oil stops flowing?
The escalation of war in Iran is already showing serious consequences for the AI world. Large investments into AI are coming from the UAE. Between many flying Dubai, the future drone attack at the oil
Polymarket Bans Nuclear Bet
You may have seen this bet. Luckily Polymarket finally decided to ban it. Turn out even a free market needs some regulation or it self-distruct. Sources: tweethttps://x.com/PolymarketStory/status/2029
Chinese labs accused of distilling Claude models
"We’ve identified industrial-scale distillation attacks on our models by DeepSeek, Moonshot AI, and MiniMax. These labs created over 24,000 fraudulent accounts and generated over 16 million exchanges
Head of AI Safety at Meta got emails nuked by OpenClaw
The Head of AI Safety at Meta.. just nuked her entire personal emails archive by giving access to her OpenClaw bot and asking and ask to remove some email. Well, the request went through the /compact