The library

research-engine · Rapid multi-source research synthesis with transparent evidence chains.

slime-rl-training · Research-grade RL training with flexible reward functions and algorithm variants.

sentry-known-pitfalls · Quick diagnosis of Sentry setup failures and error categorization issues.

research-mission-generator · Decomposing business goals into actionable research programs.

torchforge-rl-training · PyTorch-native RL training with hardware acceleration and custom loss functions.

phpstan-resolver · Migrating PHP codebases to strict type checking incrementally.

fine-tuning-with-trl · Multi-phase RLHF pipelines (SFT→Reward→PPO) where you control each alignment stage.

data-analysis-sci · When you have raw measurements and need to extract honest conclusions with proper error analysis.

verl-rl-training · Production math/reasoning tasks (GSM8K, MATH) where you need proven RL algorithms at scale.

pnpm · Monorepos needing deterministic builds, strict dependency isolation, and centralized version catalogs.

earth-life-systems · When you need to design long-term ecological studies or extract causal claims from field observations.

constitutional-ai · When you want safety alignment without human labels and need explainable reasoning in refusals.

diff-since-my-commit · Code review workflows where you need to see what others changed to your files.

experimental-design-sci · When you must isolate causal effects and design has constraints (ethics, cost, timescale).

llamaguard · Production input/output filtering where you need a specialized 7B moderation model instead of general LLM.

pubfi-dsl-server-contract · When you need strict input validation and reject SQL/unknown-field queries server-side.

history-philosophy-science · When teaching or designing science curricula and need historical/philosophical grounding.

nemo-guardrails · When you need stateful conversational guardrails beyond single-turn content moderation.

gitlab-protected-branches · Multi-team workflows where you must enforce review gates and prevent force-push accidents.

science-communication · When translating peer-reviewed findings for policy, education, or public audiences.

prompt-guard · When you need lightweight client-side jailbreak detection before sending to LLM.

scientific-method · When teaching or designing any empirical study and need a repeatable framework.

huggingface-accelerate · When you have a standard PyTorch training loop and need easy multi-GPU scaling.

citations · When generating bibliographies and need fast, accurate formatting across styles.

3d-interaction-design · When designing immersive/VR interfaces where 2D principles fail.

datenexport-portabilitaet · When implementing data portability rights (GDPR Art. 20) or migration tooling.

augmented-reality-tracking · Diagnosing why virtual AR content does not register with physical world; guides technique selection.

training-llms-megatron · Training LLMs over 1B parameters where single GPU is insufficient; achieves 47% MFU on H100.

workflow-feature-development · Starting a new feature from scratch; ensures design, planning, review, test, perf, docs all happen.

embodied-computing-and-constructionism · Teaching abstract spatial concepts (geometry, recursion) by having learners build concrete artifacts in immersive spaces.

pytorch-fsdp2 · Sharding large models across GPUs with DTensor-based parameter sharding and simpler checkpoint semantics vs FSDP1.

code-review · Post-implementation review to catch security issues, bugs, and code quality gaps before merge.

immersive-environment-design · Authoring VR/AR experiences where presence and legibility matter; ensures physical body comfort and sight-line guidance.

pytorch-lightning · Scaling training from laptop to multi-node/multi-GPU without rewriting boilerplate; automatic DDP/FSDP/DeepSpeed support.

generate-report · Creating permanent records of investigation findings with standardized naming and timestamps.

spatial-reasoning-fundamentals · Grounding 3D spatial thinking in platforms like Minecraft, VR, CAD; makes blueprint→world translation transferable.

ray-train · Running hyperparameter searches or long training runs across unstable/cloud clusters with automatic fault recovery.

world-building-block-paradigms · Designing scalable voxel worlds where modular, repeated block structures reduce manual effort.

lambda-labs-gpu-cloud · Quick provisioning of single or batch GPU instances without long-term commitment; lower cost than AWS on-demand.

rust-build · Building production Rust binaries with optimizations (LTO, codegen-units) for deployment.

beads-state · Building reactive UIs in Beads where state changes automatically propagate to views.

modal-serverless-gpu · Deploying inference-only ML models without managing containers or servers; pay-per-invocation.

done-retirement · Ensuring a clean exit where knowledge is preserved, successor is prepared, and systems are documented.

skypilot-multi-cloud-orchestration · Running large training jobs by automatically selecting cheapest cloud provider and handling spot preemption.

hook-persistence · Maintaining context across multiple hook invocations in long-running workflows without re-querying.

awq-quantization · Compressing large LLMs for edge deployment while maintaining near-FP16 accuracy.

mail-async · Multi-agent orchestration where messages must survive crashes and maintain order by timestamp

quantizing-models-bitsandbytes · Fitting 7B+ models on consumer GPUs (8-16GB VRAM) when accuracy tolerance permits <1% degradation

CA Lobby Completion Report · CA Lobby projects requiring standardized completion documentation with automated master plan progression

nudge-sync · Gastown health checks and stall detection where only the latest signal matters and fast overwrite is critical