<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0">
  <channel>
    <title>OpenTeams Engineering Blog</title>
    <link>https://openteams.com/engineering-blog/</link>
    <description>Benchmarks, systems deep dives, and hard-won lessons from OpenTeams engineers.</description>
    <item>
      <title>Parse Long PDFs Without Breaking Tables Using Unlimited-OCR-2</title>
      <link>https://openteams.com/engineering-blog/unlimited-ocr-2-long-pdf-tables/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/unlimited-ocr-2-long-pdf-tables/</guid>
      <pubDate>Mon, 02 Nov 2026 00:00:00 GMT</pubDate>
      <description>Unlimited-OCR-2 reads many PDF pages in one request, so tables split across pages come back as one table. See how it compares to page-by-page OCR.</description>
    </item>
    <item>
      <title>Your pip install Could Have Been Hijacked. Here's How It Got Fixed.</title>
      <link>https://openteams.com/engineering-blog/emudhra-pip-install-hijacked/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/emudhra-pip-install-hijacked/</guid>
      <pubDate>Fri, 30 Oct 2026 13:00:00 GMT</pubDate>
      <description>The eMudhra security incident left Python trusting a key that was the number 4. See how maintainers patched pip, certifi and conda-forge.</description>
    </item>
    <item>
      <title>What's New in Polars 2.0: Streaming by Default and Stricter Types</title>
      <link>https://openteams.com/engineering-blog/polars-2-whats-new/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/polars-2-whats-new/</guid>
      <pubDate>Mon, 12 Oct 2026 00:00:00 GMT</pubDate>
      <description>Polars 2.0 runs your lazy queries up to 2.8x faster without code changes. See the new defaults, side by side with Polars 1.x, on real data.</description>
    </item>
    <item>
      <title>How to Find Bugs in Your LLM Eval with Claude Code's build-eval</title>
      <link>https://openteams.com/engineering-blog/find-bugs-llm-eval-build-eval/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/find-bugs-llm-eval-build-eval/</guid>
      <pubDate>Wed, 07 Oct 2026 00:00:00 GMT</pubDate>
      <description>A wrong label or unfair grader can make your LLM app look better than it is. See how Claude Code's build-eval finds these problems in your eval.</description>
    </item>
    <item>
      <title>PyTorch Year in Review: OpenTeams' Contributions in 2026</title>
      <link>https://openteams.com/engineering-blog/pytorch-year-in-review-2026/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/pytorch-year-in-review-2026/</guid>
      <pubDate>Wed, 07 Oct 2026 00:00:00 GMT</pubDate>
      <description>See what our PyTorch year in review covers: a CPython-faithful Dynamo, Python 3.15 on day one, fused linear cross-entropy, MPS ops and a new build backend.</description>
    </item>
    <item>
      <title>Bare-Metal Rocky Linux Image Mode with Artifact Keeper</title>
      <link>https://openteams.com/engineering-blog/rocky-linux-image-mode-artifact-keeper/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/rocky-linux-image-mode-artifact-keeper/</guid>
      <pubDate>Tue, 06 Oct 2026 16:56:39 GMT</pubDate>
      <description>Deploy Rocky Linux image mode to bare metal with Artifact Keeper as the single source of truth for RPMs, signed bootc images, keys and Kubernetes workloads.</description>
    </item>
    <item>
      <title>Reliable Visual Regression Testing for Humans and Coding Agents</title>
      <link>https://openteams.com/engineering-blog/visual-regression-testing-jupyterlab/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/visual-regression-testing-jupyterlab/</guid>
      <pubDate>Tue, 06 Oct 2026 14:00:12 GMT</pubDate>
      <description>See how JupyterLab made visual regression testing reproducible on Linux machines, cut CI from 55 to 15 minutes and let contributors update snapshots.</description>
    </item>
    <item>
      <title>Your AI Agent Ignored AGENTS.md. Your Linter Won't Let It.</title>
      <link>https://openteams.com/engineering-blog/lint-rules-for-ai-agents/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/lint-rules-for-ai-agents/</guid>
      <pubDate>Thu, 01 Oct 2026 07:26:51 GMT</pubDate>
      <description>Use lint rules for AI agents to enforce project conventions. Custom ESLint rules like @jupyter/eslint-plugin give agents clear errors they can fix on their own.</description>
    </item>
    <item>
      <title>Where Does Jev Fit in a Software Supply Chain? We Benchmarked It Against the Open Alternatives</title>
      <link>https://openteams.com/engineering-blog/jev-bench-package-curation/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/jev-bench-package-curation/</guid>
      <pubDate>Tue, 29 Sep 2026 08:59:43 GMT</pubDate>
      <description>We benchmarked TypeSafe's Jev, Laya, CLM-8B and Claude Haiku on five package-curation triage tasks. A fine-tuned 421M model matched Jev at a tenth of the latency.</description>
    </item>
    <item>
      <title>Awesome Jupyter AI: A Map of 100+ Jupyter Extensions for AI</title>
      <link>https://openteams.com/engineering-blog/awesome-jupyter-ai-extensions/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/awesome-jupyter-ai-extensions/</guid>
      <pubDate>Fri, 25 Sep 2026 12:52:56 GMT</pubDate>
      <description>Find the right Jupyter AI extension: a curated map of 100+ chat panels, inline completers, agent bridges and MCP servers for JupyterLab and Notebook 7.</description>
    </item>
    <item>
      <title>MiniCPM5-2B quantization report</title>
      <link>https://openteams.com/engineering-blog/minicpm5-2b-quantization-report/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/minicpm5-2b-quantization-report/</guid>
      <pubDate>Wed, 23 Sep 2026 01:18:00 GMT</pubDate>
      <description>MiniCPM5-2B quantization report: the best GGUF weights and K/V cache quants on llama.cpp and BeeLlama.cpp, squeezing a SOTA model into 3 GiB RAM.</description>
    </item>
    <item>
      <title>I Asked LLMs to Review Another LLM. They Still Got It Wrong</title>
      <link>https://openteams.com/engineering-blog/llm-review-reliability/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/llm-review-reliability/</guid>
      <pubDate>Fri, 18 Sep 2026 02:24:06 GMT</pubDate>
      <description>I tested LLM reviewer models on meeting summaries to see when they catch hallucinations, when they delete true claims, and how to choose one.</description>
    </item>
    <item>
      <title>LLMs: Intelligence vs. cost</title>
      <link>https://openteams.com/engineering-blog/intelligence-vs-cost/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/intelligence-vs-cost/</guid>
      <pubDate>Wed, 02 Sep 2026 01:29:09 GMT</pubDate>
      <description>LLM intelligence vs cost: logarithmic scales in plots hide the cost chasm between frontier and cheap models; actual OpenRouter prices alter the Pareto frontier.</description>
    </item>
    <item>
      <title>pixi on UBI-micro: A Safer, Smaller Multi-Stage Container Build</title>
      <link>https://openteams.com/engineering-blog/pixi-ubi-micro-containers/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/pixi-ubi-micro-containers/</guid>
      <pubDate>Fri, 24 Jul 2026 12:21:21 GMT</pubDate>
      <description>How a UBI-micro multi-stage build shrinks a pixi container's attack surface and CVE count, with side-by-side Trivy and Grype scans on UBI full, minimal, and micro. </description>
    </item>
    <item>
      <title>The Best Code Review Says Less</title>
      <link>https://openteams.com/engineering-blog/best-code-reviews-say-less/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/best-code-reviews-say-less/</guid>
      <pubDate>Fri, 24 Jul 2026 12:06:01 GMT</pubDate>
      <description>Far too many AI code review comments are noise. Pair a human editor with AI, surface only the findings that matter, and encode that discipline in a reusable skill.</description>
    </item>
    <item>
      <title>You Don't Drive AI. You Ride It.</title>
      <link>https://openteams.com/engineering-blog/you-dont-drive-ai-you-ride-it/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/you-dont-drive-ai-you-ride-it/</guid>
      <pubDate>Thu, 09 Jul 2026 21:37:12 GMT</pubDate>
      <description>AI is stochastic, not deterministic. You don't drive it, you ride it: make documents the memory, use the variation to prototype, and verify every public claim.</description>
    </item>
    <item>
      <title>Who Is Your Code's Audience? An Engineering Team Talks AI-Written Code</title>
      <link>https://openteams.com/engineering-blog/who-is-your-codes-audience/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/who-is-your-codes-audience/</guid>
      <pubDate>Thu, 09 Jul 2026 21:36:59 GMT</pubDate>
      <description>Learn why your code's audience decides how much readability matters, as an engineering team debates AI-written code, review, and accountability.</description>
    </item>
    <item>
      <title>The tp_as_number Slot and Binary Operation Dispatch in CPython</title>
      <link>https://openteams.com/engineering-blog/tp-as-number-slot-and-binary-dispatch/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/tp-as-number-slot-and-binary-dispatch/</guid>
      <pubDate>Mon, 22 Jun 2026 14:01:40 GMT</pubDate>
      <description>How CPython dispatches binary operations via tp_as_number slots, and how TorchDynamo mirrors that model to improve correctness and reduce ad-hoc special cases.</description>
    </item>
    <item>
      <title>PDF Table Extraction: Docling vs Marker vs LlamaParse Compared</title>
      <link>https://openteams.com/engineering-blog/docling-vs-marker-vs-llamaparse/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/docling-vs-marker-vs-llamaparse/</guid>
      <pubDate>Wed, 20 May 2026 03:52:48 GMT</pubDate>
      <description>Compare three Python tools for PDF table extraction: Docling, Marker, and LlamaParse. Learn which handles merged cells and multi-level headers best.</description>
    </item>
    <item>
      <title>Sandboxing Code Mode for Local LLM Agents</title>
      <link>https://openteams.com/engineering-blog/code-mode-sandboxing-local-llms/</link>
      <guid isPermaLink="true">https://openteams.com/engineering-blog/code-mode-sandboxing-local-llms/</guid>
      <pubDate>Tue, 05 May 2026 03:01:31 GMT</pubDate>
      <description>Code mode can make local LLM agents more practical, but executing model-written code brings sandboxing back into the architecture.</description>
    </item>
  </channel>
</rss>