ContentPosts from @prashant_goyal2..
Link
@varbear shared a link, 1 week, 2 days ago
FAUN.dev()

I built a programming language using Claude Code

Cutlet usesClaude Code. The LLM emits every line. Source, build steps, and examples live on GitHub. It runs on macOS and Linux and ships aREPL. It supports arrays, strings, double numbers, a vectorizingmeta-operator, zip/filter indexing, prototypal inheritance, and a mark-and-sweepGC. Development ra.. read more  

Link
@varbear shared a link, 1 week, 2 days ago
FAUN.dev()

Cracking the Python Monorepo

Outlines a Python monorepo setup that pairsuvworkspaces withDaggerandBuildKitcaching. Builds container stages programmatically. Keeps things cache-friendly and predictable. Parsespyproject.tomland extracts the workspace graph. Copies required local packages into intermediate stages. Installs them in.. read more  

Link
@kaptain shared a link, 1 week, 2 days ago
FAUN.dev()

Running Agents on Kubernetes with Agent Sandbox

Agent Sandbox unveils the Sandbox CRD to map long-lived, singleton AI agents onto Kubernetes. It adds stable identity and lifecycle primitives. It supports runtimes like gVisor and Kata Containers. It enables zero-scale resume. It includes SandboxWarmPool with SandboxClaim and SandboxTemplate to kil.. read more  

Link
@kaptain shared a link, 1 week, 2 days ago
FAUN.dev()

Securing Production Debugging in Kubernetes

The post prescribes an on-demand SSH gateway pod. It usesshort-lived, identity-bound credentialsandKubernetes RBACto grant scoped, auditable debug sessions. It recommends anaccess brokerthat binds Roles to groups, issues ephemeral certs and OpenSSH user certificates, rotates CAs, enforces command-le.. read more  

Link
@kaptain shared a link, 1 week, 2 days ago
FAUN.dev()

RAM is getting expensive, so squeeze the most from it

The Register contrastszramandzswap. It flags a patch that claims up to 50% fasterzramops. It notes Fedora enableszramby default. It details thatzramprovides compressed in‑RAM swap (LZ4).zswapcompresses pages before writing to disk and requires on‑disk swap... read more  

RAM is getting expensive, so squeeze the most from it
Link
@kaptain shared a link, 1 week, 2 days ago
FAUN.dev()

The Invisible Rewrite: Modernizing the Image Promoter

SIG Release rewrote theimage promotercore. It cut 20% of the code. It added apipeline engine,cosignsigning, andSLSAattestations. Signing now sits separate fromsignature replication. Registry reads run in parallel - plan time dropped ~20m → ~2m. Per-request timeouts, retries, and HTTP connection reus.. read more  

Link
@kaptain shared a link, 1 week, 2 days ago
FAUN.dev()

Kubernetes v1.36 - Sneak Peek

Kubernetes v1.36 (Apr 22, 2026) enablesHPAScaleToZeroby default. That lets theHPAuseminReplicas: 0and read only controller-owned pod metrics. The release swaps long-lived image-pull secrets forephemeral KSA tokens. It deprecatesIPVS, retiresIngress NGINX, and aligns withcontainerd 2.x. The release f.. read more  

Link
@kala shared a link, 1 week, 2 days ago
FAUN.dev()

OpenClaw is a great movement, but dead product. what's next?

After talking to 50+ individuals experimenting with OpenClaw, it's clear that while many have tried it and even explored it for more than 3 days, only around 10% have attempted automating real actions. However, most struggle to maintain these automations at a production level due to challenges with .. read more  

Link
@kala shared a link, 1 week, 2 days ago
FAUN.dev()

OpenClaw Tutorial: AI Stock Agent with Exa and Milvus

An autonomous market agent ships. OpenClaw handles orchestration. Exa returns structured, semantic web results. Milvus (or Zilliz Cloud) stores vectorized trade memory. A 30‑minute Heartbeat keeps it running. Custom Skills load on demand. Recalls query 1536‑dim embeddings. Entire stack runs for abou.. read more  

OpenClaw Tutorial: AI Stock Agent with Exa and Milvus
Link
@kala shared a link, 1 week, 2 days ago
FAUN.dev()

Scaling Karpathy's Autoresearch: What Happens When the Agent Gets a GPU Cluster

A team pointedClaude Codeatautoresearchand spun up 16 Kubernetes GPUs. The setup ran ~910 experiments in 8 hours.val_bpbdropped from 1.003 to 0.974 (2.87%). Throughput climbed ~9×. Parallel factorial waves revealedAR=96as the best width. The pipeline usedH100for cheap screening andH200for validation.. read more  

Scaling Karpathy's Autoresearch: What Happens When the Agent Gets a GPU Cluster