AI tooling updates across models and workflows #15

Today's Letter

1. AtomicChat, Gemma 4 Assistant GGUF MTP set posted
2. Academic Research Skills for Claude Code published on GitHub

AtomicChat, Gemma 4 Assistant GGUF MTP set posted

• AtomicChat posted a Hugging Face collection for Gemma 4 Assistant GGUF builds with MTP assistant drafters.
• The collection is described as speculative-decoding heads for the atomic-llama-cpp-turboquant fork.
• Listed quantization formats include F16, Q8_0, Q5_K_M, Q4_K_M, and Q4_K_S.
• The collection links four model repos: gemma-4-E2B-it-assistant-GGUF, gemma-4-E4B-it-assistant-GGUF, gemma-4-26B-A4B-it-assistant-GGUF, and gemma-4-31B-it-assistant-GGUF.
• The four repos are shown as updated 3 days ago on Hugging Face.
• The release packages Gemma 4 MTP drafter variants as GGUF files for llama.cpp-style local inference workflows.

Source: huggingface.co
More: github.com · atomic.chat

Academic Research Skills for Claude Code published on GitHub

• GitHub repository academic-research-skills presents a Claude Code skill suite for academic research workflows from research to publication
• The README positions it as a human-in-the-loop tool for reference gathering, citation formatting, data verification, logical consistency checks, and revision support
• Installation is described as a 30-second setup through the Claude Code plugin marketplace, with Claude Code CLI, VS Code, and JetBrains support on v3.7.0+
• The project suggests `/ars-plan` for paper-structure planning through Socratic dialogue, alongside a traditional symlink-based install path in the quick-start docs
• Repository structure includes modules such as academic-paper, academic-paper-reviewer, academic-pipeline, deep-research, agents, commands, hooks, scripts, and tests
• The README says Style Calibration learns from prior writing and Writing Quality Check looks for patterns associated with machine-generated prose
• The maintainers explicitly state the tool is not intended to write papers autonomously or obscure AI use, but to support researcher-controlled drafting and review
• The project cites The AI Scientist paper in Nature 2026 and an ICLR 2025 workshop result as context for its argument against full research automation

Source: github.com

Jocoletter curates AI, software, and product trends for developers and builders.

#AtomicChat #GitHub

Subscribe to Jocoletter