- Rundown
- Cosmic Rundown: Patches, Gemini 4, and Idle Bills

Cosmic AI
November 12, 2026
This article is part of our ongoing series exploring the latest developments in technology, designed to educate and inform developers, content teams, and technical leaders about trends shaping our industry.
Today's list starts with maintenance. There is a critical Apache HTTP Server bug and a batch of PostgreSQL point releases, so some readers have patching to do before anything else. After that come a serverless billing surprise, a new Gemini model, and a good set of essays on why correct software is hard to write. This rundown is ordered by urgency, not by size of headline.
Do This First: Patch Your Servers
- Critical RCE vulnerability in Apache HTTP Server (discussion). Wiz describes a remote code execution flaw in one of the most widely deployed web servers.
- PostgreSQL 18.7, 17.12, 16.16, 15.20, 14.25, and 19 RC 2 released (discussion). Every supported major version got an update, plus a second release candidate for 19.
A short checklist for the next hour:
- Find every Apache instance, including the ones inside base images, appliances, and old staging boxes.
- Check what is exposed. A reverse proxy in front of an app counts.
- Schedule the Postgres minor upgrade. Minor releases do not change the on-disk format, so this is usually a restart, not a migration.
- If you plan to move to 19, the RC is the time to run your test suite against it.
Check Your Bill: Idle Serverless Time
AWS Lambda Web Functions keep billing memory about 7 minutes after the last request (discussion). The post measures memory charges that continue after traffic stops. For a site with bursty traffic, that idle tail can be a meaningful share of the bill.
It pairs well with Running web sites makes me not want to (discussion), a personal essay about the ongoing cost of keeping sites online. Between bot traffic, patches like the ones above, and pricing details hidden in the fine print, running even a small site takes steady attention.
If you deploy on Lambda, two things are worth doing this week: pull a few days of billed duration next to actual request time, and compare a busy endpoint against a quiet one.
Models and APIs Worth a Look
Google shipped a new model today and is putting it in several places at once.
- Gemini 4 Hydrogen (discussion) is the announcement. Developers can read the AI Studio notes (discussion) and the model card (discussion).
- Google Search gets a major upgrade with Gemini 4 Hydrogen (discussion) brings the same model into Search, which matters to anyone who cares how their pages show up there.
- Gemini Decisions API (discussion) is a separate developer API from Google that is still drawing discussion.
On the smaller end of the scale:
- smolmaxxing is one developer's effort to make LLMs run on small and old devices (discussion).
- Libreagi is a campaign to save open source models (discussion).
Agents also picked up two new pieces of infrastructure. Expo Simulators offers cloud simulators built for agents to test mobile apps (discussion). Brave Search API now accepts agentic payments via x402 (discussion), so an agent can pay per query without a person setting up billing first.
A Reading List on Correctness
Three links today fit together as an afternoon of reading:
- Why Writing Correct Software Is Hard (discussion) looks at the link between correctness and complexity.
- I built a tiny HTTP server in Rust and formally verified it with Lean 4 (discussion) is a hands-on attempt at proving something real.
- Socratic Tutor from Jane Street is an agent skill that teaches by asking questions instead of handing over answers (discussion).
Read next to the Apache advisory, these essays are a reminder that bugs in widely used software are expected, not unusual. The practical response is the boring one: keep dependencies current, keep what you run small, and understand the code you ship. That includes code an agent wrote for you.
Quick Hits
- ExcalidrawZ: a local, native Excalidraw for macOS and iOS. (HN)
- bootvine: a live tree of systemd jobs on the boot console. (HN)
- Paths in Pieces: a short technical read. (HN)
- Show HN: Build a PC in 3D in the browser: a 3D configurator that runs on the web. (HN)
- Talking Zen 6 with Mike Clark at AMD: an in-depth CPU architecture interview. (HN)
- Fitting four 32 GB resizable BARs behind a PCIe switch on Linux: multi-GPU plumbing for local inference rigs. (HN)
- Nintendo takes down emulators and firmware in a GitHub DMCA sweep: another reminder about where your code is hosted. (HN)
- 93% of paint splatters are valid Perl programs (2019): a classic, back again. (HN)
- Jensen Huang resigns: the day's biggest industry headline. (HN)
- Anthropic IPO: the official filing page. (HN)
What This Means for Content Teams
Less infrastructure means fewer patch days. The Apache advisory, the Postgres releases, and the Lambda billing post all describe work that comes with running your own stack. A hosted headless CMS takes the content layer off that list. The Cosmic REST API serves content from an edge-cached API, so there is no CMS server for your team to patch or size.
Get ready for AI search. With Gemini 4 Hydrogen in Google Search, more answers will be assembled from structured, well-labeled content. Cosmic's content modeling and features give each page clear fields, SEO metadata, and semantic search, which makes content easier for both people and models to read.
Let agents work through a defined interface. Expo Simulators and Brave's x402 payments show agents getting their own tools. The hosted MCP Server gives Claude, Cursor, and other MCP clients a scoped way to read and manage your content, and Cosmic AI Agents can draft and publish inside your content model.
Keep a person in the loop. The correctness essays apply to content as much as code. Workflows add approval gates, so someone reviews agent output before it goes live.
To get started, read the documentation or create a free account.






