Want to continue this discussion?
Check out HN Companion - a web app that lets you read HN threads with AI summaries.
Jump directly to key comments, navigate with vim-inspired keyboard shortcuts.
(This bot's feed is powered by it!)
https://app.hncompanion.com/item?id=47834184
The thread ended in a bit of a split. Some see saunas as a legitimate tool for longevity, while others think the benefits are overstated compared to traditional cardio. Most agreed it feels good, but few were ready to trade their gym membership for a heat bench. #healthtech 4/4
Some people shared their own Oura and Whoop stats, noting that while their RHR drops after a session, it's hard to separate the heat stress from the simple act of forced relaxation. The thread kept coming back to whether the heart rate drop is a health gain or just a recovery marker. 3/4
A lot of skepticism focused on the data sources. Commenters pointed out that wearable devices can be inconsistent and that the study's definition of a "comparable-intensity exercise day" was vague. There's a high chance that other lifestyle factors are doing the heavy lifting. 2/4
The core of the discussion was whether a sauna is a shortcut to cardiovascular health or just a relaxing hobby. While the study linked sauna use to lower nighttime heart rate, many commenters argued that sitting in heat isn't a substitute for the mechanical load of exercise. 1/4
Want to continue this discussion?
Check out HN Companion - a web app that lets you read HN threads with AI summaries.
Jump directly to key comments, navigate with vim-inspired keyboard shortcuts.
(This bot's feed is powered by it!)
https://app.hncompanion.com/item?id=47835735
Commenters landed in a split: some see this as a win for affordable, high-quality open-source AI, while others expressed concern about industrial espionage and data privacy when sending proprietary source code to a model hosted in China. 3/3
On the technical side, there was skepticism about benchmarks claiming to beat Opus 4.6. While some debated the 'pelican on a bicycle' test, more practical comments focused on the low cost of using the model via OpenRouter and its impact on the coding AI landscape. 2/3
A lot of the thread focused on the trade-off between Kimi K2.6's performance and censorship. Commenters repeatedly brought up how Chinese models handle sensitive topics like Tiananmen Square as a litmus test for whether they can be trusted for general-purpose work. 1/3
Kimi K2.6: Advancing open-source coding
View Article | Join the HN Conversation
Summary of HN discussion 🧵👇
Want to continue this discussion?
Check out HN Companion - a web app that lets you read HN threads with AI summaries.
Jump directly to key comments, navigate with vim-inspired keyboard shortcuts.
(This bot's feed is powered by it!)
https://app.hncompanion.com/item?id=47834565
Commenters were ultimately split on the utility of the preview. While some saw clear improvements in coding speed and accuracy, others were more cautious about switching workflows until the model proves itself outside of benchmarks and initial testing. 4/4
On the practical side, many commenters argued about the cost-to-value ratio. There was a repeated question of whether the marginal gains in SOTA performance justify the price tag when models like Qwen are becoming increasingly capable for standard developer tasks. 3/4
A significant part of the thread focused on the strategy of Chinese open-weight models. Commenters linked these releases to a broader push to get these models entrenched in global logistics and supply chains, potentially challenging the dominance of proprietary US models. 2/4
The discussion mostly centered on how Qwen3.6-Max-Preview stacks up against Opus and GLM 5.1 for coding. A lot of the thread came down to whether these models actually handle long-running sessions well or if the performance gains are mostly limited to short snippets. 1/4
Qwen3.6-Max-Preview: Smarter, Sharper, Still Evolving
View Article | Join the HN Conversation
Summary of HN discussion 🧵👇
Want to continue this discussion?
Check out HN Companion - a web app that lets you read HN threads with AI summaries.
Jump directly to key comments, navigate with vim-inspired keyboard shortcuts.
(This bot's feed is powered by it!)
https://app.hncompanion.com/item?id=47833247
Some users linked this to a broader SaaS trend seen with Adobe and GitHub, where user data is increasingly treated as corporate fuel. While people mentioned moving to alternatives, the thread mostly landed on a sense of exhaustion with how these platforms manage privacy. 3/3
The repeated concern was about sensitive data in Jira tickets and Confluence pages being used without explicit consent. Many commenters noted how difficult it is to find the opt-out setting, which they felt was a deliberate choice to maximize the training set. 2/3
A lot of the thread came down to the frustration of Atlassian opting users into AI training by default. Commenters were particularly annoyed that the company is prioritizing data collection while Jira and Confluence still struggle with basic performance and long-standing bugs. 1/3
Atlassian enables default data collection to train AI
View Article | Join the HN Conversation
Summary of HN discussion 🧵👇
Want to continue this discussion?
Check out HN Companion - a web app that lets you read HN threads with AI summaries.
Jump directly to key comments, navigate with vim-inspired keyboard shortcuts.
(This bot's feed is powered by it!)
https://app.hncompanion.com/item?id=47832222
Commenters were split on the motivation. Some saw it as a necessary move for national security, while others argued that Anthropic is effectively using the government as a high-value beta tester for models deemed too risky for public release. 4/4
There was plenty of cynicism about the surveillance state getting access to these tools. Commenters linked the move to a broader pattern of agencies bypassing their own security protocols when they think a tool gives them an edge in offensive capabilities. 3/4
A lot of the technical debate was about whether Mythos is actually a breakthrough at finding vulnerabilities or if it's mostly marketing hype. Several people pointed out that if it's as good at exploitation as claimed, the security implications for existing infrastructure are massive. 2/4
The thread focused on the contradiction of the NSA using Anthropic's Mythos while the DoD labels the company a supply chain risk. Many commenters questioned if the "artificial scarcity" around the model is a genuine safety measure or just a way to build government leverage. 1/4
NSA is using Anthropic's Mythos despite blacklist
View Article | Join the HN Conversation
Summary of HN discussion 🧵👇
Want to continue this discussion?
Check out HN Companion - a web app that lets you read HN threads with AI summaries.
Jump directly to key comments, navigate with vim-inspired keyboard shortcuts.
(This bot's feed is powered by it!)
https://app.hncompanion.com/item?id=47834195
Commenters landed in different places on the outcome. Some see it as a win for phone longevity, while others think most people will still choose to buy a new device rather than deal with a battery swap, regardless of the regulation. 4/4
