<h2>Must know</h2>
<ul>
<li><strong>Google releases Gemini 3.1 Pro with massive context window</strong>: Google just dropped Gemini 3.1 Pro in public preview, boasting a 1 million token context window and a 64k output limit. This new agentic and coding model scored 77.1% on the ARC-AGI-2 benchmark, a huge jump from its predecessor, and includes a configurable thinking level. It’s available across various platforms, including GitHub Copilot and Google AI Studio, and supports multimodal inputs. <a href="https://app.daily.dev/posts/PjE6Xo1XG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MySQL governance under Oracle challenged by community</strong>: Over 248 database professionals, including engineers from Percona and MariaDB, have signed an open letter urging Oracle to transfer MySQL governance to an independent non-profit foundation. Concerns include declining commit activity, a shrinking contributor base, lack of AI-era features, and an opaque roadmap. They point to PostgreSQL’s community-led success as a model, but analysts are skeptical Oracle will cede control. <a href="https://app.daily.dev/posts/GY7vcPtOu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic tightens Claude Code OAuth policy, sparks developer backlash</strong>: Anthropic updated its Claude Code documentation, restricting OAuth token usage from Pro/Max accounts in third-party tools, causing a stir among developers. While Anthropic claims it’s just a documentation cleanup, actions like a cease-and-desist to OpenClaw and banning OpenCode from OAuth access suggest a shift. Developers building commercial products are now explicitly required to use API key authentication, signaling the end of affordable flat-rate AI subscriptions for commercial development. <a href="https://app.daily.dev/feed-by-ids?id=3lsZ5Vula&amp;id=FQxtQgg9R&amp;id=GYB8zPRsi" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AI agent orchestration and memory solutions emerge</strong>: New tools like Mengram’s open-source AI memory API and Reload’s Epic platform are tackling AI agent memory and collaboration, with Mengram focusing on semantic, episodic, and procedural memory, and Reload providing shared long-term memory for AI coding teams. <a href="https://app.daily.dev/feed-by-ids?id=0tYkAozTe&amp;id=MQxvV1b9u" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI infrastructure demands drive massive power plant project</strong>: SoftBank’s SB Energy plans a $33 billion, 9.2 gigawatt natural gas power plant in the Ohio-Kentucky border region, potentially the largest in the US, to meet the surging energy demands of AI data centers, though it faces environmental concerns and a long development timeline. <a href="https://app.daily.dev/feed-by-ids?id=IYsqAR4Cy&amp;id=VBQEBjah6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding tools reshape developer workflows and productivity</strong>: AI-assisted coding is driving shifts in developer technology choices, with TypeScript gaining traction due to better AI tool performance, and dual-agent workflows emerging to address single AI agent limitations, though human judgment remains critical. <a href="https://app.daily.dev/feed-by-ids?id=2LocixQWJ&amp;id=Gzw0VkhBv&amp;id=IHGn3gHuu&amp;id=SEDqLFz7t&amp;id=Tw5ITz4bj&amp;id=VadEPVBil&amp;id=XlyWxUnez" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub Copilot integrates code referencing, Heroku updates AI models</strong>: GitHub Copilot now highlights matching code snippets from public repositories with links to sources and licenses, while Heroku’s Managed Inference and Agents add-on expands its model offerings to include Claude Opus 4.6 and Sonnet 4.6, deprecating older versions. <a href="https://app.daily.dev/feed-by-ids?id=kcB1LfSD2&amp;id=lOJYvaow4&amp;id=rsFWrvtAL&amp;id=vDLyK99UX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI and robotics industrial adoption accelerates</strong>: Toyota is deploying Agility Robotics Digit humanoid robots in its RAV4 plant for auto parts unloading, following a year-long pilot, while NVIDIA’s survey shows 90% of telecom respondents report AI increasing revenue and reducing costs, indicating a broader industrial shift. <a href="https://app.daily.dev/feed-by-ids?id=WnnenuOfb&amp;id=p1OVwpWSR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tailscale workload identity federation now generally available</strong>: Tailscale’s workload identity federation is now GA, allowing CI/CD pipelines and cloud workloads to authenticate using short-lived OIDC tokens, eliminating static credentials and integrating with Terraform and Kubernetes operators. <a href="https://app.daily.dev/posts/ekx9IIs1v" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vitest outperforms Jest in performance benchmarks</strong>: Vitest 3.x shows significant performance gains over Jest 30, with 5.6x faster cold starts and 28x faster watch mode re-runs in a production monorepo, attributed to its Vite-based transform pipeline and shared module caching. <a href="https://app.daily.dev/posts/T3G4VUzdk" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI slop floods open source, maintainers drown in garbage</strong>: Apparently, AI coding tools have made code generation ‘nearly free and near-instant,’ which sounds great until you realize it’s flooding open-source projects with ‘AI slop.’ InfoWorld reports maintainers are now overwhelmed by pull requests that look professional but are riddled with subtle errors, security flaws, or unmaintainable complexity. So much for AI making everyone’s lives easier; now we just have more garbage to sift through. <a href="https://app.daily.dev/posts/Ixa8rNXuv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Wix engineering finds AI productivity is mostly a feeling</strong>: Wix Engineering ran a 5-week experiment with 86 developers using AI coding tools. While users of Cursor ‘felt’ more productive, the data showed almost no measurable lift. Claude Code did show gains, but only after developers stopped micromanaging it. So, it seems AI can make you feel good, but actual productivity might just be a placebo effect, or you’re just doing more work to get the AI to do its job. <a href="https://app.daily.dev/posts/qQhctEUPB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent demos fail in production, who could have seen that coming?</strong>: Aditya Godbole, speaking at Fifth Elephant Pune 2026, points out that AI agent demos often fail in production because ‘unbounded telemetry data’ overwhelms context windows. He suggests schema-driven knowledge graphs as a ‘reasoning substrate’ for production AI agents. Because, you know, the problem isn’t that we’re trying to make AIs do everything, it’s that we haven’t built a sufficiently complex Rube Goldberg machine to manage their inevitable data indigestion. <a href="https://app.daily.dev/posts/dtY8IcMQN" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
