Topic
Everything filed under Infrastructure, newest first.
RSS · JSON · All topics
The efficient frontier of LLM inference
Philip Kiely explains the efficient frontier of LLM inference, separating techniques that trade off cost, latency, and quality (such as quantization) from techniques like speculative decoding and disaggregation that improve overall system performance.
5 min · 1,166 words
Please Implement A Quorum Database
Transcript of an HPTS 2026 talk arguing that leaderless, quorum-based replication is practical for production databases today and deserves more use than Raft-style leader replication, walking through the design space, prior systems, and the techniques that make quorums work.
16 min · 3,776 words
Protomaps: Self-hosting map tiles has gotten cheap and easy
Bruce Röttgers walks through self-hosting map tiles with Protomaps' PMTiles format, which serves a single file over HTTP Range requests: extracting a region with the CLI, deploying it to Bunny.net storage and CDN, and embedding the map, as a cheap alternative to commercial map APIs.
3 min · 767 words
When One GPU Isn't Enough: Serving GLM-5.3 Across Two NVIDIA DGX Stations with vLLM
NVIDIA shows how to serve the frontier-scale GLM-5.3 model locally across two DGX Stations linked by ConnectX-8 RDMA with vLLM, reaching about 607.8 tokens per second.
18 min · 4,077 words
Critical RCE Vulnerability in Apache HTTP Server: Everything You Need To Know
Wiz explains CVE-2026-119618, a critical race condition in Apache HTTP Server's HTTP/2 module (mod_http2) that could let a remote unauthenticated attacker corrupt heap memory for RCE, with affected versions and mitigation such as restricting protocols to HTTP/1.1.
2 min · 461 words
Why we chose our own compute over AWS Lambda Web Functions
Easel tested AWS Lambda Web Functions for running web apps and found they kept billing memory for roughly 5–7 minutes after the last request, so it chose its own compute, which also benchmarked faster on the same Next.js app.
5 min · 1,063 words
An Open Field Guide to Inference Engineering
Modelplane's open field guide for platform engineers new to inference, covering how models and GPUs work together, running inference in production, the tooling landscape, and how to grow into an inference engineering role, with figures as of October 2026.
39 min · 8,965 words
How a false smoke alarm almost drowned Jane Street's only NYC datacenter
A war story from Jane Street's systems team: a false smoke alarm, a datacenter with no cooling, and a sprint to pull every power cord before the sprinklers went off.
7 min · 1,611 words
Val Town founder Steve Krouse announces the company is winding down after four years, pausing billing, linking migration guides to Cloudflare, Vercel, Netlify and Render, and setting a timeline that ends with open-sourcing Val Town on Feb 1, 2027.
1 min · 265 words
Xata announces it is joining ClickHouse to bring instant copy-on-write branching, scale-to-zero and anonymized production data to ClickHouse Managed Postgres, aimed at AI agents and per-PR databases; existing Xata customers keep running as usual.
6 min · 1,392 words
One Stolen Vendor Key, Every Customer's Data Exposed: Inside the Qubole Supply Chain Attack
Sentra describes a supply chain attack in which stolen Qubole credentials were used to reach data at dozens of Qubole customers through cross-account access they had once granted the vendor, and how it was uncovered in under 24 hours.
9 min · 2,073 words
Fun things to do with your VM/370 CE machine
Ricardo Bánffy's hands-on guide to the VM/370 Community Edition mainframe OS: what CE adds over IBM's last unrestricted VM/370, running it with Docker, 3270 terminals, logging on, writing BASIC, FORTRAN, PL/I and REXX programs, navigating disks, uploading files and games, adding users, editors, networking and printing.
25 min · 5,729 words
Update on October 2026 security incident
Fireworks AI updates its October 2026 security bulletin: an LLMJacking-linked threat actor accessed account and billing information and, for some accounts, secrets, datasets and incomplete model weights, partly via a breached third-party cloud provider; Fireworks details remediation with Palo Alto Networks Unit 42.
3 min · 588 words
Canada launches the Real-Time Rail, bringing 24/7 payment infrastructure to the national economy
Payments Canada announces the launch of Canada's Real-Time Rail (RTR), a national payment system that lets participating financial institutions, credit unions and payment service providers exchange, clear and settle payments in real time, 24/7, with data-rich ISO 20022 messaging.
13 min · 2,917 words
Our newest TPUs: 8tx and 8i Max, built for the agentic era
Google expands its eighth-generation TPU portfolio with TPU 8tx, aimed at secure inference, and TPU 8i Max, built for massive context windows, as it scales infrastructure for agentic AI workloads.
8 min · 1,829 words
What I run on my $10 USD OpenStick homelab
Laurens Bosscher walks through the self-hosted services he runs on a $10 OpenStick LTE dongle with 512 MB of RAM and four Cortex-A53 cores, listing each container and its memory footprint, from reverse proxy and monitoring to bookmark archiving and update notifications.
5 min · 1,077 words
Creating Custom Agents: Inside Bronto's Alert Investigation Agent
Bronto details how its open-source alert investigation agent was built with LangGraph and LangChain, and how custom agents pair with observability tools for faster incident response.
10 min · 2,252 words
How Stream Moderates Live Video & Audio
Stream explains the architecture behind moderating live video and audio calls: detection pipelines, policy enforcement, and how moderation fits into real-time media infrastructure.
8 min · 1,775 words
From Heroku to Convox, and why we built Rack Gateway
DocSpring recounts moving off Heroku onto Convox and building Rack Gateway, an access-control gateway aimed at SOC 2-friendly infrastructure access.
7 min · 1,656 words
Outlook.com Is Quietly Rejecting Email From Domains Without DMARC
MailRoute reports that since November 5, 2026, Outlook.com, Hotmail, Live and MSN have been bouncing some mail from domains with no DMARC record, including low-volume senders, and explains what the bounce looks like and why a single v=DMARC1; p=none TXT record fixes it.
4 min · 1,025 words