<h2>Must know</h2>
<ul>
<li>Google releases Gemini 3.1 Pro, challenging competitors: Google has launched Gemini 3.1 Pro in public preview, positioning it as their most powerful agentic and coding model. It boasts a 1 million token context window 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 metric. The model is available across various platforms, including GitHub Copilot and Google AI Studio, with pricing set lower than some rivals, according to Google’s announcement. <a href="https://app.daily.dev/posts/PjE6Xo1XG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Anthropic’s Claude Code policy shift causes developer uproar: Anthropic updated its Claude Code documentation, restricting OAuth token usage from Pro/Max subscription accounts in third-party tools, sparking community backlash. While Anthropic claims it’s a documentation cleanup, official legal documents explicitly state these tokens are for Claude Code and <a href="http://Claude.ai" target="_blank" rel="noopener nofollow">Claude.ai</a> only, requiring API keys for developers building products. This move, coupled with actions against tools like OpenClaw, suggests a shift away from affordable flat-rate AI subscriptions for commercial development, as reported by multiple sources. <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>
<li>AI agents reshape software development, but new challenges emerge: Agentic AI is fundamentally changing software development, moving from prompt-response to autonomous agents that plan and execute. Tools like Kiro and Claude Code are boosting individual productivity, but this speed introduces new challenges like ‘AI slop’ in open-source projects and subtle production bugs that traditional code reviews miss. The role of developers is evolving towards agent orchestration and ensuring human oversight remains central to maintaining code quality and managing new bottlenecks in CI/CD pipelines, as highlighted by Wix Engineering’s experiments and industry experts. <a href="https://app.daily.dev/feed-by-ids?id=037a5YIjx&amp;id=4CjaWv3pP&amp;id=7XxAvY7Df&amp;id=Bl6akHr3q&amp;id=DqRWRtC6t&amp;id=F8eGl7g0G&amp;id=LCg5ueVO7&amp;id=SBASpAo7A&amp;id=UJqBwwcg1&amp;id=Uw1m3wdti" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li>Hugging Face launches community evals for AI model benchmarking: Hugging Face introduced ‘Community Evals,’ a beta feature aimed at creating transparent, community-driven AI model benchmarking by allowing benchmark datasets to host leaderboards and aggregate evaluation results. <a href="https://app.daily.dev/posts/9WqEmBh9G" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AI memory and agent management solutions emerge: Mengram launched an open-source AI memory API with semantic, episodic, and procedural memory types, while Reload introduced Epic, an AI workforce management platform providing shared long-term memory for AI coding agents. <a href="https://app.daily.dev/feed-by-ids?id=0tYkAozTe&amp;id=MQxvV1b9u" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>MySQL governance under Oracle challenged by community: Over 248 database professionals signed an open letter urging Oracle to transfer MySQL governance to an independent non-profit foundation, citing declining commit activity, a shrinking contributor pool, and lack of AI-era features. <a href="https://app.daily.dev/posts/GY7vcPtOu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AI infrastructure development and energy demands surge: Cisco and NVIDIA launched the Secure AI Factory, a modular architecture for AI compute, networking, storage, and security, while SoftBank plans a massive 9.2 gigawatt natural gas power plant to meet AI data center energy needs. <a href="https://app.daily.dev/feed-by-ids?id=IYsqAR4Cy&amp;id=VBQEBjah6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AI and robotics industrial adoption accelerates: Toyota deployed seven Agility Robotics Digit humanoid robots at its RAV4 plant for auto parts unloading, and NVIDIA’s survey reports 90% of telecom respondents see AI increasing revenue and reducing costs. <a href="https://app.daily.dev/feed-by-ids?id=WnnenuOfb&amp;id=p1OVwpWSR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AI agent security, memory, and performance challenges: Azazel, an open-source eBPF-based runtime tracer, captures kernel-level events for AI agents, while Anthropic’s Sonnet 4.6 shows a ‘split personality’ in benchmarks, excelling in agentic coding but regressing on one-shot tasks. <a href="https://app.daily.dev/feed-by-ids?id=cbOVi24pM&amp;id=dtY8IcMQN&amp;id=iGzH52Vzz&amp;id=k9XThzl85&amp;id=qBKhvArRq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Moderne adds Python support to OpenRewrite refactoring platform: Moderne now supports Python in its OpenRewrite code refactoring platform, enabling automated refactoring recipes for Python codebases to upgrade APIs, manage dependencies, and enforce formatting across polyglot environments. <a href="https://app.daily.dev/posts/KHjOD9xYZ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li>AI productivity: feeling productive vs actually being productive: Wix Engineering’s experiment found that while developers using Cursor felt more productive, there was almost no measurable productivity lift. Meanwhile, Claude Code showed significant gains, but only after developers stopped micromanaging it. So, it seems our AI tools are great at making us feel busy, but actual results require us to trust the machine, or maybe just use a different one. Who knew? <a href="https://app.daily.dev/posts/qQhctEUPB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AI content authenticity: Microsoft’s blueprint, industry’s shrug: Microsoft’s AI safety team proposed a blueprint for verifying online content authenticity, evaluating 60 combinations of provenance tracking and watermarking. Yet, Microsoft’s chief scientific officer wouldn’t commit to implementing these across their own platforms. Experts note the approach is promising but can’t verify factual accuracy, and industry adoption faces headwinds from engagement-driven business models. So, we have a solution, but nobody wants to use it because clicks are more important than truth. Classic. <a href="https://app.daily.dev/posts/SCrxseqRQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AI’s impact on jobs: the eternal debate continues: A European study found AI increases labor productivity by 4% with no short-term job losses, while an Indian think tank says AI isn’t an immediate threat to their IT sector. Yet, 27% of Indian firms expect headcount to shrink due to AI. Meanwhile, some Silicon Valley ‘Chicken Littles’ are still arguing AI won’t destroy jobs, just amplify human capability. It’s almost like nobody actually knows what’s going to happen, but everyone has a strong opinion. <a href="https://app.daily.dev/feed-by-ids?id=iZlTkI8Me&amp;id=tiBubXcsb&amp;id=yOsJJlumD" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>