» curated · synthesized
Skip the noise.
Read the signal.
Curated tech news and synthesis for developers and technology professionals.
» Latest posts
194 postsPorting Gemma-4 (2B/4B/12B) to AWS Inferentia2: A Field Report
How Gemma-4 was ported to AWS Inferentia2 by bypassing vLLM/NxD, solving mixed attention heads and KV-sharing via direct model tracing.
A Memory-Bounded HTML-to-PDF Architecture for Bursty Workloads
How a solo-built PDF service uses a Postgres-backed queue and streaming Chromium renders to cap memory use regardless of document size.
Hooks that stop AI coding agents from claiming unverified reads
A hook-based kit blocks AI coding agents from editing unread files, enforcing scoped changes, consumer checks, and verified session handoffs.
CommitBrief — AI code reviews, right in your terminal
A provider-agnostic, local-first CLI that reviews your staged changes, a historic range, or a whole GitHub pull request. Zero telemetry, no server. Free and open source.
commitbrief.comLLM Inference Latency: Why Memory Bandwidth Beats FLOPS
Why LLM inference speed depends on GPU memory bandwidth, not FLOPS. Real tok/s benchmarks across GPUs, quantization levels, and batch sizes.
Context Engineering: Why AI Agents Need More Than Good Prompts
AI agent output depends less on prompt wording and more on curated context. Why tool design, memory, and context curation now beat prompt engineering.
Forge3D Spaces: Text-to-3D Building CAD Running in the Browser
Forge3D Spaces converts text prompts into 3D buildings in-browser using an LLM-to-solver pipeline, WebGPU fallback, and real-time CSG wall cutting.
CodeGraph Turns Repos into Queryable Graphs to Predict Breakage
CodeGraph converts GitHub repos into a Neo4j knowledge graph, using GraphRAG and a blast-radius score to predict code change impact.
4 self-hosting failures that quietly report success
Four silent self-hosting failures — 200 OK pages, false healthchecks, empty backups, vanished WoL packets — and how to actually catch them.
How a Second Middleware Silently Broke Inngest's TypeScript Types
A deep dive into why stacking two Inngest middlewares silently collapsed step.run's TypeScript type to {} — and the actual root-cause fix.
kotlin-lib-mcp: an MCP server exposing real sources of any Maven library to AI
kotlin-lib-mcp is an open-source MCP server letting AI assistants read real Maven library sources, API surface, and KDoc directly.