AI users are invited to share which subscriptions, models, coding tools, and agent platforms they currently use. The questions aim to capture how AI workflows and tool choices have changed over the past year. (news.ycombinator.com)
Google may use Gemini chat history from personal accounts to train its models, including for subscribers, when Keep Activity is enabled. Disabling it stops that use but deletes chats after 72 hours, while some human-reviewed chats may be retained for up to three years. (news.ycombinator.com)
The daily digest
Today's best Hacker News stories, summarized and screenshotted, one email a day.
A Hacker News poster asks which practical AI skills individuals can learn online to benefit economically from LLMs and agents. They want specific, hands-on recommendations suited to someone with a moderately technical background, without requiring a degree. (news.ycombinator.com)
The daily digest
Today's best Hacker News stories, summarized and screenshotted, one email a day.
An AI agent user asks how people preserve context, decisions, and client-specific rules across sessions and different agents. They’re also exploring whether to build a tool to reduce repeated explanations and costly mistakes. (news.ycombinator.com)
The daily digest
Today's best Hacker News stories, summarized and screenshotted, one email a day.
Local LLM users are asked to share the hardware they rely on for everyday work. The question welcomes both cost-effective setups and powerful home servers. (news.ycombinator.com)
AI has yet to independently produce anything genuinely new, the author argues, relying instead on human insight or existing information. The question asks readers to name a clear example of AI creating something novel on its own. (news.ycombinator.com)
Opus 5.5 is producing noticeably worse responses and losing track of tasks, according to a user testing it across several agents and projects. The decline reportedly began within the past half day. (news.ycombinator.com)
AI could displace workers, or its capabilities could be exaggerated as pre-IPO hype. The discussion asks what people believe about AI’s impact and promises. (news.ycombinator.com)
Claude Code users report being prompted to log in, with the process hanging on claude.ai. Several commenters say they can no longer use the coding assistant. (news.ycombinator.com)
A programmer says relying on AI tools has made work feel less creative and enjoyable than hands-on coding. They ask whether others miss the flow and satisfaction they once found in their jobs. (news.ycombinator.com)
Claude Opus 5.5 users report recent declines in performance, including coding errors and ignored instructions. They criticize Anthropic for potentially changing the model without changing its name or notifying users. (news.ycombinator.com)
Tab-completion coding can still help developers generate predictable code chunks and review them as they go. The question asks whether specialized autocomplete models are still being developed or have been eclipsed by autonomous coding tools. (news.ycombinator.com)
Many people develop AI projects privately and consider sharing ideas to inspire others or incorporate feedback. Examples include a Linux desktop picture viewer by date, highlighting common needs in file management tools. (news.ycombinator.com)
People are discussing how they evaluate AI agent skills before and during use. Methods include systematic testing or simply reading the skill instructions and trying them out. (news.ycombinator.com)
ChatGPT offers superior insights compared to Codex but has limited tool calling capabilities, according to a user. They want to combine both models' strengths locally on their laptop for better AI agent performance. (news.ycombinator.com)
Coding agents produce more code than developers can thoroughly review for quality and robustness. Knowledge of clean code and software architecture helps developers assess agent-written code. (news.ycombinator.com)
Widely accessible AI offers high intelligence, but people have been slow to use it to change the world. The question is whether human imagination, rather than intelligence, has become the bottleneck. (news.ycombinator.com)
A reader asks how the US midterm election results could affect AI. They seek views on open-source AI, data centers, and space-based computing. (news.ycombinator.com)
LLMs often use “it’s not A, it’s B” to draw clear distinctions, and the pattern is also common in human writing. Commenters attribute its prevalence to online text and later training, while noting that prompting can change the style. (news.ycombinator.com)
A software engineer worries that rapidly improving AI tools will erode advantages in cybersecurity and production troubleshooting. They ask which skills can help engineers stay valuable as AI agents take on more complex work. (news.ycombinator.com)
Anthropic has launched a research preview of Claude Duet, an AI agent that assists with tasks and manages attention. A quote from The Matrix highlights concerns about AI potentially shaping or dominating human civilization. (news.ycombinator.com)
A custom agent harness lets a developer safely edit project files without installing Claude or relying on its permissions model. The developer uses the small tool daily and asks why others build and keep using their own harnesses. (news.ycombinator.com)
Jev and decision models cost about ten times as much as DeepSeek, prompting questions about their value. The user asks what specialized models offer beyond sending DeepSeek the same input with strict formatting instructions. (news.ycombinator.com)
An Ask HN poster wonders what share of Hacker News submissions and comments are AI-generated, guessing it could exceed 50%. They say their own posts and comments were typed by hand. (news.ycombinator.com)
Claude users are invited to share the funniest apps they built with leftover quota before a reset. The Hacker News post has no replies yet. (news.ycombinator.com)
Personal assistants may replace some apps as people delegate tasks to AI models or agents. Direct requests can be easier than navigating apps and dashboards. (news.ycombinator.com)
OpenAI will remove the apparently unrestricted high-quality chat mode from Plus in December. The change could simply make existing invisible limits visible rather than substantially reducing access. (news.ycombinator.com)
AI coding tools have not emptied bug backlogs or cleared feature requests, despite claims that they can replace programmers. Developers still have more work than these tools have eliminated. (news.ycombinator.com)
A Kimi Code user says two coding tasks exhausted the $19 plan’s five-hour quota after about $1.37 in API-equivalent usage, without completing either task. They question whether the restrictive limit reflects a quota-accounting bug or an intentional subscription cap. (news.ycombinator.com)
Open weights models like Ornith-1.5 have largely disappeared, with recent releases becoming infrequent. Developers now appear more focused on monetization rather than consistent open-source releases. (news.ycombinator.com)
A request seeks recommendations for AI tools that create pitch decks and visuals. No specific tools or answers are included. (news.ycombinator.com)