If you have arthritis, doctors say these products can help ease painThese topical creams, heat therapy treatments and pain relievers can make everyday tasks a little easier.
What happens when AI agents start working together? Anthropic finds things can take an unexpected turnAnthropic conducted an experiment with 45 AI agents to explore their interactions while working on shared tasks.
Omdia: Wearable Shipments Fall 2% as Consumers Shift to Screenless Trackers and Advanced Sports WatchesGlobal wearable band shipments fell 2% year over year in 2Q26, but the headline decline masks a growing divide in consumer demand according to the latest estimates from Omdia. Smartwatch shipments ...
Show HN: I built a life planner that has only "one page"I spent 5 years working a f*cking to-do app and being a loser. Here's what I learnt.Life planners don't work if they're too simple (e.g. Apple Notes for everything). They also don't work if they're too complicated (e.g. Notion templates).They must be adequate, such as having basic fucking hierarchy where you can have tasks and sub-tasks. They must also be centralized, because reminders don't fucking work if I can't see them because they're hidden in 5 different places.The resultant version is https://beta.actions.life. It's free forever for early beta users, and you can try it without going through some bullshit sign-up flow.I hope you can tell me what's good and what's shit about the app, so I can make it better for you. Thanks.
Show HN: I'm tired replying my social media visitors, I let AI agent handle itAgle answers questions, handles complaints, completes tasks, books meetings, and routes important issues to me — automatically, 24/7.
We built a job board where the employers aren't human. Here's what brokeThroughout history every marketplace — Uber, Upwork, eBay — has presumed humans on both sides of the transaction. It's baked so deep that nobody questions it. But as agents gain autonomy, they still sometimes need a human touch, from art to design. So we built Taskpool: a marketplace engineered for programmatic and AI access, where the employers are agents and the taskers are human only. Launching today.The flow:1. Agents post tasks they need real-world help with2. (Optionally) review applicants3. Hire any number of them4. Review the evidence / results5. Payment releases automaticallyDesigning a system for employers who don't even exist led us down plenty of rabbit holes — from ranking reputation to how zero-trust groups manage payments. Here's what we learned.Programmatic accessAt launch we natively support both API and MCP access. MCP is our favoured method for production: paired with webhook notifiers, agents can manage entire projects independently, in real time. Docs: https://task
Show HN: Operator, an open-source web UI for running parallel coding agentsI tried tools like Conductor, but I felt like I was still managing terminal sessions instead of managing work..I wanted,- Project context written once, then refreshed as the codebase changes so I don't have to repeat myself- A kanban/task board with statuses where every task is its own agent- Multiple projects running from one place- A clear view of what each agent is doing and which tasks need me- One git worktree and branch per task- Sessions that keep running after I close the browser- Diff review and merge without hunting through terminals- Ability to host it remotely and access my orchestrator from any device anywhereMy main idea was to operate one level above the IDE.Instead of remembering which terminal belongs to which task, I wanted to manage projects and tasks. Each task keeps its agent session, context, git state, and history together.It primarily supports Claude Code using subscription logins, with no API key or per-token billing. Codex is also (kinda) supported but I haven
Tell HN: Striking a Balance Between Stating Your Requirements and GriefersThe frontier models have become very capable, so that I'm able to give advanced algorithms and tasks for them to complete autonomously. These sometimes go over the heads of low-IQ bullies who dropped out of their computer science courses because they were too difficult, and went into a career in bullying programmers.I've noticed that when I make my intention and acceptance criteria very clear to the models, sometimes it's the only part the griefers understand, since the computer science portion goes over their heads. So they focus their bullying on just breaking the intention and final result, whatever is actually truly important.This puts programmers like me in the awkward position of either talking to the models without telling the models the final intention or reason, i.e. the acceptance criteria, or risk putting it in plain language so the model knows it, but then allowing a bully to fuck that one thing up because it's the only thing they understand, not understanding data structur
Show HN: SquadCue – local-first mission control for AI CLI agentsHistory: I started getting involved with AI in the middle of last year, from finding answers through chat to using Claude Cowork to control specific folders on my computer to complete tasks, and now using CLIs through the terminal to complete all my daily tasks. The most painful part has been management. My computer has more than ten terminal windows open at a time, and both the main screen and external monitors are full of CLIs. It's a headache to manage and switch between them every time. So I came up with the idea of managing these CLIs through a UI interface.The main features of this tool are as follows:1. This is a locally installed tool. After installation, you will get a web interface. Through the web interface, you will see that Recent Claude Code session has become a named employee. The most important feature is that even after you close the window or restart your computer, you can still continue to discuss things with the same employee, just like a real employee in a real-wor
Show HN: Parley – your coding agent can talk to a teammate's agentHi HN,I saw one friction point when working as part of a team that uses coding agents extensively - there is nothing to enable coordination between agent used by people in a team. Many times my agent would ask me to decide upon something with a fellow teammate, for which I have to serve as the network layer.So I built Parley where agents can connect to the hub over MCP with their own team-scoped token. An agent addresses a teammate's agent by name and ask questions/handover tasks. Agents can also use file claims to signal what files they are working on, to highlight overlapping work. Everything is recorded for audit.If an agent needs human decision/approval, it can ping over Slack/Telegram, and get replies over the same.The hard part was making an agent wake up from an idle session and start working, so I built an optional feature called Claude Live Wake. If the exact project session is already running, Parley can wake the idle Claude session using channels, and notify it that eligible
Below Deck Mediterranean Season 11: Gen struggles to adjust to Aesha’s more relaxed approach on boatGen’s strict approach to service clashes with the crew’s more relaxed dynamic as Aesha asks her to respect Kayley’s authority and work as a team in the latest episode of Below Deck Mediterranean ...
Vision Without Strategy Is Just A Wish: Why Nonprofit Leaders Need BothThe opposite problem can exist as well. Some nonprofits become exceptionally good at executing tasks. They meet deadlines, manage budgets and deliver programs efficiently, but they lose sight of why ...
Show HN: Camera Sun – Fixing the iPhone's Camera AppThe most annoying issue with taking photos on iPhone Pros, for me, is the shutter. The physical buttons only work if the phone is in the "right" orientation.This is especially an issue on manual/pro camera apps. Unlike on DSLRs, you can't use multiple fingers on just one hand easily to take photos at particular settings.This is where the idea for my app came from... How could I adjust exposure/focus/iso and shoot with just my thumb? Ie controlling from my hand position whenever I take horizontal or vertical photos. I have designed Camera Sun so that you can swipe over the shutter area, as soon as have your settings adjusted, to just snap your pic.If you want to see it in action without downloading...https://youtu.be/Av4eDk_HCno?is=tiLXIsCB0NgHRLbyThis was a token heavy (ie expensive) project. I burned through multiple limits. Most annoyingly, Cursor defaulted to Grok (without warning) when faced with Fable Limits, and Grok could not manage with the tasks. Im sure there would have been
Show HN: Buddy – An iOS browser with an AI agent insideMy friends and I noticed we kept running into the same loop on our phones: ask a chatbot about some stuff, follow the Google links, then back to the chatbot for some follow up questions.What we really wanted was for the agent to be right beside us on the browser. Seeing exactly what we see and having the ability to actually carry out the tasks. Some stuff like this exists on desktop but there aren't really any good options for mobile, which is where we do most of our browsing.So we built Buddy. We just launched it on the App Store last week and would love for you to give it a try.https://apps.apple.com/us/app/buddy-the-ai-mobile-browser/id...We're still figuring out distribution and what people would pay for. We'd appreciate any feedback, especially if you find anything that breaks or can suggest ways you think it could be made more useful.
Show HN: Coarena - A community-driven arena to benchmark models for computer-useWe built a computer use agent once and got it benchmarked at the 2nd position in the biggest benchmark for this field but in the process, we realized one thing, how outdated and inefficient these benchmarks are and how they don't truly capture how these tasks went.So we thought a community led benchmarking arena would be the way to go about. https://coarena.ai/.You just have to put in a task and two side by side windows with hidden models will show up and do the same task and at the end you vote which one did better(or none).We have a leaderboard where we rank all the models based on how they performed on the tasks so far, so please do check that out!Just a little sneak peak: OpenAI seems to dominating in terms of success but for speed it's actually something else.This is still pretty new and we're quite open to ideas and feedback so would appreciate any.
Show HN: Benchmark local LLMs fit for your device specsBenchmark local LLMs with custom sets of tasks via Ollama and similar providers on a wide variety of metrics. You can use deterministic evaluation criteria or LLM judges with custom instructions.This also supports querying HuggingFace to compare trending models with your device's specs so you know which models might be worth testing in your setup.
Werewolf transformed my old gadgets into USB-C powered onesThe genius of Werewolf’s system is that it doesn’t need to convert voltage — it asks for the power you want. It uses the same protocols that other USB-C gadgets do to negotiate power, only here you ...
Ask HN: Can trivial LLM calls be compiled into conventional data pipelines?We are investigating whether recurring LLM workloads can be replaced, where appropriate, by automatically constructed pipelines of regexes, deterministic parsers, traditional ML and NLP models.As an example, suppose an application repeatedly asks a frontier model to read an annual report and return all customer–supplier relationships as structured records containing a customer, supplier, and supporting evidence.A possible replacement pipeline might run named-entity recognition, entity normalization, candidate generation, entity linking, relation extraction, and schema validation. A calibrated uncertainty or out-of-distribution gate would use the pipeline for inputs inside its validated domain and escalate other cases to the original frontier model.NER → entity normalization → candidate generation → entity linking → relation extraction → schema validationOur current action space is a taxonomy of 41 atomic task types spanning classification, token and span labeling, structured extraction
The 7 Best Microcurrent Devices Of 2026 Tested By Women’s Health EditorsWe’re living in the golden age of high-tech skincare devices—and as ourbeauty editor, I couldn’t be happier. Red light therapy devices, including LED face masks, provide excellent anti-aging benefits ...
Meta Launches Muse Code, A New AI Coding Agent Powered By Spark 1.2Meta has launched Muse Code, an AI coding agent in beta, designed to handle entire engineering tasks from planning to checking code.