LodeHQSubscribe →

ClickHouse ray tracer, Kubelet leak fix, CF 4x speed

Dev · 2026-07-01

Languages & Frameworks
ClickHouse SQL Powers a Full‑Scene Ray Tracer4 MIN

The repo implements a full path tracer using only ClickHouse SQL, outputting a PNG image without any UDFs or external code. By treating each pixel as a row and leveraging ClickHouse’s parallel query engine, it shows that sophisticated graphics pipelines can run in a pure SQL environment, opening doors for data‑centric compute workloads.

Tools & Platforms
Box3D launches as open‑source 3D physics engine to bridge gaps in game physics7 MIN

Box3D is the new open‑source 3D physics engine born from Box2D, adding triangle‑mesh, height‑field and large‑world support while keeping a lightweight C17 core. It targets game developers frustrated with existing middleware, offering SIMD‑accelerated contacts, multithreading hooks, and continuous collision out of the box.

Run Kubernetes Clusters Directly in Your Browser with Webernetes13 MIN

Webernetes, an open‑source TypeScript rewrite, lets you spin up a full‑featured Kubernetes cluster inside a web page. It supports pod lifecycles, DNS, deployments and a browser‑based image registry, all without WebAssembly. This makes cluster demos, education and rapid prototyping instantly accessible without any server install.

CloudFormation Express mode slashes deployment time up to 4x3 MIN

AWS CloudFormation Express mode cuts deployment latency by up to fourfold, finishing as soon as resources are configured instead of waiting for stabilization checks. This lets developers and AI‑assisted tools get immediate feedback, speeding iteration cycles and reducing idle time in production rollouts.

Kubelet memory leak in Kubernetes 1.36 traced to context cancellations and patched6 MIN

A 2 GiB DigitalOcean node on Kubernetes 1.36 ran out of memory as the kubelet process grew unchecked. Using Go's pprof, the author traced the leak to a flood of context cancellation objects and applied a targeted fix that stops the leak, stabilizing the node and preventing pod restarts.

AWS Certificate Manager adds ACME, enabling fully automated public TLS6 MIN

ACM now runs an ACMEv2 server so any compatible client, Certbot, cert‑manager, acme.sh, can issue public certificates from Amazon Trust Services. Centralized IAM controls and domain scoping let PKI teams enforce policies and audit usage, cutting manual renewal risk as cert lifetimes shrink.

GitHub’s Stacked PRs let you split massive changes into review‑friendly layers2 MIN

Stacked PRs let developers break a large diff into a chain of focused pull requests that each build on the previous one. The GitHub UI shows a stack map and the optional gh‑stack CLI handles cascading rebases, so reviewers stay contextual and merges happen in one click. Teams ship faster with fewer conflicts.

Gemma 4 12B removes encoders to enable fast on‑device multimodal AI3 MIN

Google unveiled Gemma 4 12B, a dense multimodal model that throws out separate vision and audio encoders and feeds raw visual/audio data straight into its LLM backbone. This encoder‑free design cuts latency and memory use, making high‑performance AI run locally on consumer devices.

Git 2.55 adds incremental multi‑pack indexes to speed huge repos15 MIN

Git 2.55 lets the repack command write incremental multi‑pack index (MIDX) chains, so large repositories can add new pack layers without rewriting the whole index. This cuts maintenance write overhead and speeds object lookup, a boon for massive codebases on GitHub and elsewhere.

Google sunsets Tenor API, forcing X, Discord and WhatsApp to find new GIF source1 MIN

Google announced today that the public Tenor GIF API will be shut down on June 30, 2026. The move cuts off third‑party access for apps like X, Discord, WhatsApp and Bluesky, pushing them to adopt alternative services, while Google‑owned products keep using Tenor internally.

AI-Assisted Development
fenic turns AI‑augmented queries into reusable DataFrame pipelines10 MIN

fenic is a DataFrame engine that embeds LLM operators, extract, classify, summarize, embed, directly into query pipelines. It lets engineers and AI agents author, inspect, and reuse typed, cacheable workflows instead of ad‑hoc prompts or regex scripts, turning exploratory analysis into durable artifacts.

Get Dev in your inbox, every issue.
Subscribe free
Privacy · Terms · About · Contact
© 2026 LodeHQ