Google Open Source

@opensource.google

Announcing new open source releases, exploring projects, sharing how we approach FOSS, and supporting communities around the world.

@opensource.google

The past few days have been a blur. Open Source Summit + Embedded Linux Conference Europe was packed with news and information. What open source tidbits have you the most excited?

@opensource.google

Open source is free as in puppies, not free as in beer. 🐶📦

While research often looks at upstream maintenance, @sophiaiv.bsky.social studied the effort to USE open source software at scale across Google.

What makes updates take days vs. hours? Dive into the data: goo.gle/cost-of-oss

Google Open Source Research card 01/04: "How Much Does It Cost to Use Open Source Software?" Subheadline: "Open source is 'free as in puppies, not free as in beer.' Beyond upstream maintenance, we studied the engineering effort required to use and update OSS packages at Google." Features portrait of Sophia Vargas, Google Open Source, with two bullets: "Evaluates software engineering hours required to update OSS packages within Google" and "Shares observational findings, complexity indicators, and study limitations." Study Overview box: "How long do package updates take—and which metrics predict relative effort?" Footer: Inside the post: Research questions, update complexity observations, study scope, and limitations. goo.gle/cost-of-ossGoogle Open Source Research card 02/04: "What the Study Examined & Observed in Practice." Three observational cards: (1) Research Question 1 (RQ1) — "<= 4 Hours vs. Long Tail: How Long Do Package Updates Take? We evaluated how long it takes an engineer (not an upstream maintainer) to update an OSS package at Google: most take 4 hours or less, while a long tail takes days to months." (2) Research Question 2 (RQ2) — "Stronger Effort Indicators: What Predicts Update Complexity? Results suggested local patches and customization, number of package dependencies, and number of upstream contributors were better indicators of update complexity than other metrics evaluated." (3) Observed Comparison — "Inconclusive Results: Package Age & Internal Usage Profile: Analysis did not yield any conclusive results regarding package age or how many teams were using that package internally. While upstream breaking changes can significantly impact update complexity, this was not detectable by package age or version alone."Google Open Source Research card 03/04: "Survey Context & Study Limitations." Top callout: "2024 Employee Survey Chart (346 Respondents): Top-Ranked Characteristics Making Updates Take Longer — 1st: Google-specific patches or customization (49%); 2nd: Upstream breaking changes (37%); 3rd: Google breaking changes (31%)." Four-box grid of study limitations: (1) Business Priority Sampling: Designed around existing data to minimize the impact to engineering roadmaps. (2) Did Not Dictate Tooling: Did not specify tools (including genAI features) or isolate specific tool impacts. (3) Sample Bias Towards Go: Data had significant bias towards Go; future studies encouraged on language ecosystems. (4) Only Active Projects: Focused on active packages.Google Open Source Research card 04/04 for Bluesky (@opensource.google): "Explore the Approach & Full Research Report." Top box ("What's in the Blog Post & Research Paper"): Author Sophia Vargas (@sophiaiv.bsky.social), Google Open Source; Paper: "How much does it cost to maintain a copy of an OSS package?"; Contents: Study design and overview of topline results; Scope: Observational study of internal OSS package updates. Middle box ("Research Questions Explored"): 1. RQ1: How long does it take for an engineer (not an upstream maintainer) to update an open source software package at Google? 2. RQ2: Can complexity metrics predict the relative effort or amount of time required to update a package at Google? CTA: Link in post above: https://goo.gle/cost-of-oss
@opensource.google

We checked in with the authors! When the team evaluated options and made the decision back in January 2026, Pyrefly was further along in maturity and closer to a stable release for our internal scale, which was the primary deciding factor.

@opensource.google

Join Google at Open Source Summit + Embedded Linux Conference Europe in Prague! 🇨🇿 Swing by our Booth #31 to: ✨ Catch live demos 🤝 Meet the team 🎁 Grab exclusive gear We can't wait to connect with the community! 🚀

@opensource.google

🚀 Don't miss Erin McKean's Open Source Summit keynote!

Discover the history & future of @Docsy, why docs matter EVEN MORE in the age of AI, and how everyone (and their AI agents!) can contribute.

📍 Session: Technical documentation for everyone (And Their AI Agents, Too)

sched.co/2Rasq

A promotional graphic for the Linux Foundation's Open Source Summit Europe and Embedded Linux Conference Europe, taking place on 7-9 October 2026 in Prague, Czechia. The theme of the event is "CODE. COMMUNITY. COLLABORATION." The banner announces Erin McKean as a Keynote Speaker, listing her title as "Staff Developer Relations Engineer, Google Open Source Programs Office, Google." A circular headshot in the bottom right corner shows Erin, a smiling woman with glasses and short, graying wavy hair. The overall design features a dark blue background with orange, peach, and white typography, accented by stylized botanical illustrations in the corners.
@opensource.google

This Week in Open Source #18 is live! 🌐✨ 🧮 Inside Gemma 4 12B inference & 100% local RAG (LiteRT + Qdrant) 🧪 Adaptive agent training w/ EnvHarness 🏗️ @mozilla.org & Mila's open AI stack ⚠️ The maintainer cost of LLM "vibe-coding" 🗓️ Oct/Nov events: goo.gle/twios-18 #OpenSource

Google Open Source banner: This Week in Open Source #18 (Oct 2, 2026). Headline: What happens when open source AI moves from benchmarks to infrastructure? Featuring Gemma 4 local RAG, EnvHarness agent sandboxes, and Mozilla & Mila's open stack. Card 2 of 4: Open Source AI & Infrastructure highlights. Features Gemma 4 12B KV-cache sizing math and offline LiteRT+Qdrant RAG, Google's EnvHarness adaptive RL agent sandboxes, and Mozilla & Mila Quebec's sovereign open AI ecosystem. Card 3 of 4: Community & Sovereignty highlights. Features maintainer costs of unreviewed AI "vibe-coding" PRs, UK Government open-source digital sovereignty stewardship, and native FreeBSD-to-Linux syscall translation on PlayStation 5. Card 4 of 4 (Bluesky): Call to action pointing to goo.gle/twios-18 in post above. Highlights reads from @mozilla.org, Mila, @qdrant.to, LiteRT, and Gemma 4, plus Oct/Nov events including ValkeyConf, MCP Dev Summit, and OSS Europe.
Older posts