<h2>Must know</h2>
<ul>
<li><strong>Google releases powerful new Gemini 3.1 Pro model</strong>: Google has launched Gemini 3.1 Pro in public preview, touting it as their most powerful agentic and coding model yet. It boasts a 1 million token context window, multimodal inputs, and scored 77.1% on the ARC-AGI-2 benchmark, significantly outperforming its predecessor and competitors like Anthropic’s Opus 4.6 and OpenAI’s GPT-5.2 on this specific test. Pricing is also competitive at $2 per million input tokens, making it an interesting option for AI-powered development. <a href="https://app.daily.dev/posts/PjE6Xo1XG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MySQL community demands independent governance from Oracle</strong>: Over 248 database professionals, including engineers from major players like Percona and MariaDB, have signed an open letter urging Oracle to transfer MySQL’s governance to an independent non-profit foundation. They cite declining commit activity, a shrinking contributor base, lack of AI-era features, and an opaque roadmap under Oracle’s stewardship. This move aims to address growing fragmentation and ensure MySQL’s future relevance, drawing parallels to PostgreSQL’s community-led success. <a href="https://app.daily.dev/posts/GY7vcPtOu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic’s Claude Code policy shift causes developer uproar</strong>: Anthropic updated its Claude Code documentation, restricting OAuth token usage from Pro/Max subscription accounts in third-party tools, leading to community backlash. While Anthropic claims it’s just a documentation cleanup, developers are concerned this signals the end of affordable flat-rate AI subscriptions for commercial development. Official legal documents now explicitly state that API key authentication via Claude Console is required for developers building products, impacting tools like OpenClaw and NanoClaw. <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 security 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 to fuel AI data centers, highlighting the immense energy needs of the burgeoning AI industry. <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 workflows and raise quality concerns</strong>: AI-assisted coding is driving shifts in developer technology choices and workflows, with tools like Claude Code showing significant output gains, but also raising concerns about ‘AI slop’ in open-source projects and the need for better production visibility to catch subtle bugs. <a href="https://app.daily.dev/feed-by-ids?id=2LocixQWJ&amp;id=Gzw0VkhBv&amp;id=IHGn3gHuu&amp;id=IR0A3b0XK&amp;id=SEDqLFz7t&amp;id=Tw5ITz4bj&amp;id=VadEPVBil&amp;id=XlyWxUnez&amp;id=2Xamq3Lpe" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI and robotics accelerate industrial adoption</strong>: Toyota is deploying Agility Robotics’ Digit humanoid robots for auto parts handling, and NVIDIA’s survey shows 90% of telecom respondents report AI increasing revenue and reducing costs, indicating a strong trend of AI and robotics moving into industrial deployment. <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>Hugging Face launches community evals for AI model benchmarking</strong>: Hugging Face introduced ‘Community Evals,’ a new beta feature aimed at making AI model evaluation more transparent and community-driven by allowing benchmark datasets to host leaderboards and aggregate evaluation results from model repositories. <a href="https://app.daily.dev/posts/9WqEmBh9G" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New tools for AI agent control and device interaction</strong>: Callstack’s open-source <code>agent-device</code> CLI tool allows AI agents to programmatically control iOS and Android devices, while <code>Trajectory Explorer</code> aims to provide better visibility into AI agent decision-making during execution. <a href="https://app.daily.dev/feed-by-ids?id=Ekfc0OFB7&amp;id=AzPQg0qJn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Apache Polaris graduates to top-level project</strong>: Apache Polaris, an open-source catalog for Apache Iceberg and data lakehouse interoperability originally from Snowflake, has graduated to a Top-Level Project under the Apache Software Foundation, signaling mature governance and momentum for open data lakehouse standards. <a href="https://app.daily.dev/posts/ovRl52TUb" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will make you productive, but only if you pay more</strong>: AI coding tools, initially promising unlimited access for a mere $20/month, have apparently shifted to usage-based billing, with some $200 tiers still hitting rate limits. So, you can be super productive, but only if you’re willing to shell out significantly more cash, and even then, don’t expect truly ‘unlimited’ access. What a surprise. <a href="https://app.daily.dev/feed-by-ids?id=dz3RmQVnk&amp;id=j5Q6D03Mc&amp;id=opnnf1fiE&amp;id=uNeZBI42S" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI search’s hardest problem is ‘intent understanding,’ not LLMs</strong>: Meilisearch claims that ‘intent understanding’ is the hardest unsolved problem in AI-powered search, not LLM integration. Apparently, decomposing natural language queries into structured search parameters is rocket science, and every team building conversational search is just reinventing the wheel. Maybe if the LLMs were actually good at understanding intent, this wouldn’t be such a ‘hard’ problem. <a href="https://app.daily.dev/posts/9RuTrwk1b" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will save you time and money, if you believe this guy</strong>: One individual claims six AI tools (Perplexity, Claude, Whisper Flow, Cursor, Feedly, Granola) reduce their daily work to about two hours. While it’s great for them, this sounds suspiciously like the kind of anecdotal evidence that fuels unrealistic expectations for everyone else. Your mileage, and actual work output, may vary significantly. <a href="https://app.daily.dev/posts/X2M941duB" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
