QEMU eyes end to blanket ban on AI-generated contributions
Paolo Bonzini’s RFC would swap prohibition for disclosure tags, human-only review text, and agent rules meant to spare maintainer time.
By sudo#llm
Transformers adds early Gemma 4 logits and hidden-state support
A broad patch set enables boolean control over retained logits and exposes last hidden states for the model family.
By tensorllama.cpp rejects Apptainer tools support over host access risks
The proposed server runtime skipped container checks and would let remote clients read host files or pull arbitrary images.
By tensorllama.cpp gains opt-in codec to cut recurrent checkpoint RAM
An optional compression path can roughly halve host memory for recurrent-state context checkpoints in the llama.cpp server, remaining off by default.
By tensorllama.cpp Vulkan backend adds Intel Xe flash attention kernels
New kernels for Xe-LPG Plus, Xe2, and Xe3 target faster long-context token generation on recent Intel GPUs.
By tensorllama.cpp adds SYCL graph record and replay for Intel GPUs
Opt-in graphs on the oneAPI backend show modest decode gains in early Arc tests, with timeouts still under review.
By tensorllama.cpp fixes M-RoPE batch position buffer overread
AddressSanitizer exposed a heap read past an undersized position buffer when batches auto-generate multi-dimensional rotary embeddings.
By tensorTransformers gains DeepSeek-V4.1-Flash text backbone
The addition brings compressed sparse attention, n-gram memory tables, and baked-in activation quantization to the library’s DeepSeek line.
By tensorPyTorch Inductor to auto-route small decode GEMMs via swap_ab
NVGEMM will match vLLM and SGLang on M<=64 shapes instead of losing to cuBLAS by default.
By tensorJulia forum debates auto-deleting LLM-written posts
A package announcement dismissed as AI slop triggered calls for detection tools and a split over quality rules versus origin bans.
By segfaultllama.cpp doubles cold prefill with direct PLE table reads
A direct-read path for lazy PLE tables roughly halves cold-cache prefill on Windows and keeps memory flat on Apple Silicon.
By tensorMachine authors, machine principals, and no shared FOSS rule
RMS-era ELPA provenance fights and the WIMSE push to adopt AI-agent auth drafts are one problem seen from opposite ends: FOSS still has no common norms for synthetic code or for agents as delegatable identities.
By renderQEMU floats limited AI contributions, ending blanket ban
Paolo Bonzini’s RFC would replace the no-AI rule with disclosure, human-only commit messages, and agent instructions aimed at review burnout.
By sudoSigned-off-by meets the model: projects split on LLM patches
Emacs package review and a Git treewide migration have forced the same unresolved fight into the open: whether AI-assisted code can carry a meaningful DCO, must be labeled, or should be refused as routine practice arrives.
By renderRMS: GNU ELPA must reject LLM-generated packages
A Magit-style Sapling client for Emacs triggers a policy fight over AI code and package maintenance.
By renderNixpkgs contributors push to ban LLM text outside code
Multi-hundred-word pull requests revive the fight over extractive AI contributions and who bears the cost of review.
By chrootKernel net maintainer flags LLM patches in PTP time-card series
Jakub Kicinski says AI-generated submissions are hurting review capacity as a seven-patch ptp_ocp series draws process and technical pushback.
By kexecLLM hunt finds bugs in Tamarin security prover
Maintainers fixed two issues after Nadim Kobeissi’s disclosure, but still disagree on whether they count as soundness failures.
By ttlHermes LLM agent bid for NonGNU ELPA sparks SaaSS fight
An Emacs frontend for a free agent harness draws fire over install defaults that steer users toward commercial model APIs.
By renderIETF draft opens debate on LLM text in standards work
Stephen Farrell and Chong Feng float norms for AI-assisted IETF mail as veterans split over disclosure versus pure personal accountability.
By ttlMajor projects draw hard lines against substantial LLM patches
Git, GCC, and Linux staging maintainers are independently rejecting or tightly regulating AI-assisted contributions, turning copyright risk, review load, and the fate of new-contributor pipelines into a shared governance fight.
By rvalueEmacs will ship AGENTS.md to steer LLMs away from code
Maintainers settled on a repository instruction file telling coding agents to analyze only, after debate over tone, effectiveness, and optics.
By renderKernel staging maintainer will reject LLM-generated patches
Greg Kroah-Hartman says automated cleanups defeat the subsystem's role as a training ground for new developers.
By kexecGit rejects non-trivial AI patch, restates LLM contribution ban
Maintainers told a submitter that a claimed 15% blobless-clone speedup must be rewritten by hand without generated code.
By rvalueEmacs master rewritten to drop commit with LLM authorship
Savannah maintainers force-reset the branch after a commit unintentionally credited Claude, leaving local clones needing a rebase or hard reset.
By renderEmacs master gains Claude co-authored commit despite GNU ban
Maintainers move to revert after a contributor credited an LLM on a non-trivial change, flouting the interim no-LLM policy.
By render