My computers were connected to an UPS surge projector. Various kitchen machinery and extension sockets that was not doesn't work anymore. The neighbor's TV blew.
There was a bright white flash inside of my living room, just under an open window and directly above the my UPS surge protector and then all my devices turned off including the surge projector. The PSU in one of the computers is gone, but not the motherboard and rest.
I was sitting on the concrete floor in the middle of the room watching it happen.
I live in Thailand in the middle of a big flat coconut groove with only two houses. Except for the palm trees, my house is the tallest structure in a 200 x 200 meters (300 feet)+ area and we have frequent huge thunder storms, but this one was in a category on its own. It made my concrete walls tremble. It was so intense. Spectacular show by nature.
Only two months ago, I bought a socket outlet tester. It reported that my sockets was not grounded. I got an electricians out to fix that. I suspect this is the reason why the lightning chooses me. We have frequent big thunderstorms and haven't had this issue before, but now there is a ground connection.
I don't know if there is a proper grounding rod. There is a cable going from the house into the ground, but I don't know what it is attached too. The rest of the electric installation is amateur's business so could be connected to nothing. The house is a long-term rental and the electric installation came with it, but I can change it regardless of it's is a rental.
The electricity doesn't even come directly from the power company's cables. It goes in to the owners house, where the meter is. From there they have pulled a 200 meter cable into ours. There is a fuse box in my house.
Any suggestions on what to do next? A big metal stick somewhere in the garden?"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["compute"],"value":"Ask HN: My house just got hit by lightning and it fried the PSU in my computer"}},"_tags":["story","author_TowerTall","story_31270349","ask_hn"],"author":"TowerTall","children":[31270509,31270564,31270586,31270601,31270615,31270632,31270633,31270638,31270653,31270679,31270685,31270691,31270694,31270754,31270864,31270987,31271156,31271402,31271552,31271990,31272071,31274070,31274338,31279488,31280930],"created_at":"2022-05-05T06:47:29Z","created_at_i":1651733249,"num_comments":79,"objectID":"31270349","points":74,"story_id":31270349,"story_text":"If everything is wired correctly in my house, should all equipment attached to my UPS have survived this with no impact or is a direct lightning strike so powerful that even high-quality home-grade UPS must give up?
My computers were connected to an UPS surge projector. Various kitchen machinery and extension sockets that was not doesn't work anymore. The neighbor's TV blew.
There was a bright white flash inside of my living room, just under an open window and directly above the my UPS surge protector and then all my devices turned off including the surge projector. The PSU in one of the computers is gone, but not the motherboard and rest.
I was sitting on the concrete floor in the middle of the room watching it happen.
I live in Thailand in the middle of a big flat coconut groove with only two houses. Except for the palm trees, my house is the tallest structure in a 200 x 200 meters (300 feet)+ area and we have frequent huge thunder storms, but this one was in a category on its own. It made my concrete walls tremble. It was so intense. Spectacular show by nature.
Only two months ago, I bought a socket outlet tester. It reported that my sockets was not grounded. I got an electricians out to fix that. I suspect this is the reason why the lightning chooses me. We have frequent big thunderstorms and haven't had this issue before, but now there is a ground connection.
I don't know if there is a proper grounding rod. There is a cable going from the house into the ground, but I don't know what it is attached too. The rest of the electric installation is amateur's business so could be connected to nothing. The house is a long-term rental and the electric installation came with it, but I can change it regardless of it's is a rental.
The electricity doesn't even come directly from the power company's cables. It goes in to the owners house, where the meter is. From there they have pulled a 200 meter cable into ours. There is a fuse box in my house.
Any suggestions on what to do next? A big metal stick somewhere in the garden?","title":"Ask HN: My house just got hit by lightning and it fried the PSU in my computer","updated_at":"2024-09-20T11:06:31Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"kylequest"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["thunder","compute"],"value":"123456\npassword\nphpbb\nqwerty\n12345\njesus\n12345678\n1234\nabc123\nletmein\ntest\nlove\n123\npassword1\nhello\nmonkey\ndragon\ntrustno1\n111111\niloveyou\n1234567\nshadow\n123456789\nchrist\nsunshine\nmaster\ncomputer\nprincess\ntigger\nfootball\nangel\njesus1\n123123\nwhatever\nfreedom\nkiller\nasdf\nsoccer\nsuperman\nmichael\ncheese\ninternet\njoshua\nfuckyou\nblessed\nbaseball\nstarwars\n000000\npurple\njordan\nfaith\nsummer\nashley\nbuster\nheaven\npepper\n7777777\nhunter\nlovely\nandrew\nthomas\nangels\ncharlie\ndaniel\n1111\njennifer\nsingle\nhannah\nqazwsx\nhappy\nmatrix\npass\naaaaaa\n654321\namanda\nnothing\nginger\nmother\nsnoopy\njessica\nwelcome\npokemon\niloveyou1\n11111\nmustang\nhelpme\njustin\njasmine\norange\ntesting\napple\nmichelle\npeace\nsecret\n1\ngrace\nwilliam\niloveyou2\nnicole\n666666\nmuffin\ngateway\nfuckyou1\nasshole\nhahaha\npoop\nblessing\nblahblah\nmyspace1\nmatthew\ncanada\nsilver\nrobert\nforever\nasdfgh\nrachel\nrainbow\nguitar\npeanut\nbatman\ncookie\nbailey\nsoccer1\nmickey\nbiteme\nhello1\neminem\ndakota\nsamantha\ncompaq\ndiamond\ntaylor\nforum\njohn316\nrichard\nblink182\npeaches\ncool\nflower\nscooter\nbanana\njames\nasdfasdf\nvictory\nlondon\n123qwe\n123321\nstartrek\ngeorge\nwinner\nmaggie\ntrinity\nonline\n123abc\nchicken\njunior\nchris\npassw0rd\naustin\nsparky\nadmin\nmerlin\ngoogle\nfriends\nhope\nshalom\nnintendo\nlooking\nharley\nsmokey\n7777\njoseph\nlucky\ndigital\na\nthunder\nspirit\nbandit\nenter\nanthony\ncorvette\nhockey\npower\nbenjamin\niloveyou!\n1q2w3e\nviper\ngenesis\nknight\nqwerty1\ncreative\nfoobar\nadidas\nrotimi\nslayer\nwisdom\npraise\nzxcvbnm\nsamuel\nmike\ndallas\ngreen\ntesttest\nmaverick\nonelove\ndavid\nmylove\nchurch\nfriend\ngod\ndestiny\nnone\nmicrosoft\n222222\nbubbles\n11111111\ncocacola\njordan23\nilovegod\nfootball1\nloving\nnathan\nemmanuel\nscooby\nfuckoff\nsammy\nmaxwell\njason\njohn\n1q2w3e4r\nbaby\nred123\nblabla\nprince\nqwert\nchelsea\n55555\nangel1\nhardcore\ndexter\nsaved\n112233\nhallo\njasper\ndanielle\nkitten\ncassie\nstella\nprayer\nhotdog\nwindows\nmustdie\ngates\nbillgates\nghbdtn\ngfhjkm\n1234567890"},"title":{"matchLevel":"none","matchedWords":[],"value":"Pony password stealer own password list"},"url":{"matchLevel":"none","matchedWords":[],"value":""}},"_tags":["story","author_kylequest","story_6857010","ask_hn"],"author":"kylequest","created_at":"2013-12-05T20:32:27Z","created_at_i":1386275547,"num_comments":0,"objectID":"6857010","points":2,"story_id":6857010,"story_text":"123456\npassword\nphpbb\nqwerty\n12345\njesus\n12345678\n1234\nabc123\nletmein\ntest\nlove\n123\npassword1\nhello\nmonkey\ndragon\ntrustno1\n111111\niloveyou\n1234567\nshadow\n123456789\nchrist\nsunshine\nmaster\ncomputer\nprincess\ntigger\nfootball\nangel\njesus1\n123123\nwhatever\nfreedom\nkiller\nasdf\nsoccer\nsuperman\nmichael\ncheese\ninternet\njoshua\nfuckyou\nblessed\nbaseball\nstarwars\n000000\npurple\njordan\nfaith\nsummer\nashley\nbuster\nheaven\npepper\n7777777\nhunter\nlovely\nandrew\nthomas\nangels\ncharlie\ndaniel\n1111\njennifer\nsingle\nhannah\nqazwsx\nhappy\nmatrix\npass\naaaaaa\n654321\namanda\nnothing\nginger\nmother\nsnoopy\njessica\nwelcome\npokemon\niloveyou1\n11111\nmustang\nhelpme\njustin\njasmine\norange\ntesting\napple\nmichelle\npeace\nsecret\n1\ngrace\nwilliam\niloveyou2\nnicole\n666666\nmuffin\ngateway\nfuckyou1\nasshole\nhahaha\npoop\nblessing\nblahblah\nmyspace1\nmatthew\ncanada\nsilver\nrobert\nforever\nasdfgh\nrachel\nrainbow\nguitar\npeanut\nbatman\ncookie\nbailey\nsoccer1\nmickey\nbiteme\nhello1\neminem\ndakota\nsamantha\ncompaq\ndiamond\ntaylor\nforum\njohn316\nrichard\nblink182\npeaches\ncool\nflower\nscooter\nbanana\njames\nasdfasdf\nvictory\nlondon\n123qwe\n123321\nstartrek\ngeorge\nwinner\nmaggie\ntrinity\nonline\n123abc\nchicken\njunior\nchris\npassw0rd\naustin\nsparky\nadmin\nmerlin\ngoogle\nfriends\nhope\nshalom\nnintendo\nlooking\nharley\nsmokey\n7777\njoseph\nlucky\ndigital\na\nthunder\nspirit\nbandit\nenter\nanthony\ncorvette\nhockey\npower\nbenjamin\niloveyou!\n1q2w3e\nviper\ngenesis\nknight\nqwerty1\ncreative\nfoobar\nadidas\nrotimi\nslayer\nwisdom\npraise\nzxcvbnm\nsamuel\nmike\ndallas\ngreen\ntesttest\nmaverick\nonelove\ndavid\nmylove\nchurch\nfriend\ngod\ndestiny\nnone\nmicrosoft\n222222\nbubbles\n11111111\ncocacola\njordan23\nilovegod\nfootball1\nloving\nnathan\nemmanuel\nscooby\nfuckoff\nsammy\nmaxwell\njason\njohn\n1q2w3e4r\nbaby\nred123\nblabla\nprince\nqwert\nchelsea\n55555\nangel1\nhardcore\ndexter\nsaved\n112233\nhallo\njasper\ndanielle\nkitten\ncassie\nstella\nprayer\nhotdog\nwindows\nmustdie\ngates\nbillgates\nghbdtn\ngfhjkm\n1234567890","title":"Pony password stealer own password list","updated_at":"2024-09-19T20:14:13Z","url":""},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"pritambarhate"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["thunder","compute"],"value":"Hi,
While reading the BlackBlaze Compute Story I came across, Packet's ARM Offering. (https://www.packet.net/bare-metal/servers/compute/)
Their c1.large.arm offers ARM Cavium ThunderX with 96 Physical Cores running at 2.0 Ghz.
Has anybody used such a configuration for Web Application Servers? Possibly something running Go?
Since many web applications need high concurrency and wait on DB queries, I think having these large number of cores might be beneficial.
Any thoughts and inputs on this will be well appreciated."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["thunder"],"value":"Ask HN: ARM Cavium ThunderX (96 Physical Cores) for Web Application Servers?"}},"_tags":["story","author_pritambarhate","story_16752184","ask_hn"],"author":"pritambarhate","children":[16752453],"created_at":"2018-04-04T05:22:53Z","created_at_i":1522819373,"num_comments":1,"objectID":"16752184","points":2,"story_id":16752184,"story_text":"Hi,
While reading the BlackBlaze Compute Story I came across, Packet's ARM Offering. (https://www.packet.net/bare-metal/servers/compute/)
Their c1.large.arm offers ARM Cavium ThunderX with 96 Physical Cores running at 2.0 Ghz.
Has anybody used such a configuration for Web Application Servers? Possibly something running Go?
Since many web applications need high concurrency and wait on DB queries, I think having these large number of cores might be beneficial.
Any thoughts and inputs on this will be well appreciated.","title":"Ask HN: ARM Cavium ThunderX (96 Physical Cores) for Web Application Servers?","updated_at":"2024-09-20T02:15:22Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"flatroze"},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["compute"],"value":"ARM Desktop Tower Computer"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["thunder"],"value":"https://www.avantek.co.uk/store/avantek-32-core-cavium-thunderx-arm-desktop.html"}},"_tags":["story","author_flatroze","story_16069989"],"author":"flatroze","created_at":"2018-01-04T13:00:01Z","created_at_i":1515070801,"num_comments":0,"objectID":"16069989","points":2,"story_id":16069989,"title":"ARM Desktop Tower Computer","updated_at":"2024-09-20T01:54:49Z","url":"https://www.avantek.co.uk/store/avantek-32-core-cavium-thunderx-arm-desktop.html"},{"_highlightResult":{"author":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["thunder"],"value":"ill_thinker"},"story_text":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["compute"],"value":"Today I was told by my manager that my arguments come across jumbled and incoherent.
This is not news to me. Since a young age I've found it incredibly difficult to reason about and form an argument for or against a topic. I first encountered this in my high school design and technology exam when I was asked to articulate why I would choose a particular material over another. No matter how much I tried, I simply could not reason about the different materials and which one would be better.
12 years on and I now have a PhD in computer science. I find software development extremely straightforward and the accompanying decision making easy. But during design discussions before development takes place, I still find it very difficult to present a reasoned argument as to why I'd design a piece of software one way over another. I mostly rely on an internal feeling that either provides positive or negative emotions (intuition). Project level decisions are also affected by this.
The absence of reasoned argument is not due to a lack of trying. When I attempt to reason about a problem I struggle to think of the questions I should be asking. Thoughts about a topic do not come naturally, and often they are very cyclic and disjointed. It is only through writing down my thoughts that my argument gains clarity. You can imagine how problematic this thought process is for active oral discussions, my discourse appears sporadic and poorly thought-out.
Is there anything I can do to improve my reasoning skills so I can participate coherently in oral discussions?"},"title":{"matchLevel":"none","matchedWords":[],"value":"Ask HN: How do I get better at creating well-reasoned arguments?"}},"_tags":["story","author_ill_thinker","story_24703551","ask_hn"],"author":"ill_thinker","children":[24703695,24703703,24705516,24709341,24710104,24710920,24712980,24713235,24714340,24728782],"created_at":"2020-10-06T23:14:28Z","created_at_i":1602026068,"num_comments":9,"objectID":"24703551","points":16,"story_id":24703551,"story_text":"Today I was told by my manager that my arguments come across jumbled and incoherent.
This is not news to me. Since a young age I've found it incredibly difficult to reason about and form an argument for or against a topic. I first encountered this in my high school design and technology exam when I was asked to articulate why I would choose a particular material over another. No matter how much I tried, I simply could not reason about the different materials and which one would be better.
12 years on and I now have a PhD in computer science. I find software development extremely straightforward and the accompanying decision making easy. But during design discussions before development takes place, I still find it very difficult to present a reasoned argument as to why I'd design a piece of software one way over another. I mostly rely on an internal feeling that either provides positive or negative emotions (intuition). Project level decisions are also affected by this.
The absence of reasoned argument is not due to a lack of trying. When I attempt to reason about a problem I struggle to think of the questions I should be asking. Thoughts about a topic do not come naturally, and often they are very cyclic and disjointed. It is only through writing down my thoughts that my argument gains clarity. You can imagine how problematic this thought process is for active oral discussions, my discourse appears sporadic and poorly thought-out.
Is there anything I can do to improve my reasoning skills so I can participate coherently in oral discussions?","title":"Ask HN: How do I get better at creating well-reasoned arguments?","updated_at":"2025-10-06T21:07:41Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"weakgeekinseek"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["thunder","compute"],"value":"i am an entrepreneurial sharp business mind in search of a technologically savvy counterpart. i have many ideas including a previous sold at in excess of $200 million. i am a thinker with a knack for financial gains but very limited to my software development knowledge. i am looking for a computer wiz to help develop my ideas and form partnerships for all possible ventures possible with your skills. my email is 150mikkimase@gmail.com and my cell is 6467097799. please contact me, Mikki, at any time suitable for you. i look forward to your inquiries and am excited to move forward with my new partner(s)."},"title":{"matchLevel":"none","matchedWords":[],"value":"Entrepreneur searching for developer counterpart"},"url":{"matchLevel":"none","matchedWords":[],"value":""}},"_tags":["story","author_weakgeekinseek","story_9620290","ask_hn"],"author":"weakgeekinseek","created_at":"2015-05-28T18:54:42Z","created_at_i":1432839282,"num_comments":0,"objectID":"9620290","points":1,"story_id":9620290,"story_text":"i am an entrepreneurial sharp business mind in search of a technologically savvy counterpart. i have many ideas including a previous sold at in excess of $200 million. i am a thinker with a knack for financial gains but very limited to my software development knowledge. i am looking for a computer wiz to help develop my ideas and form partnerships for all possible ventures possible with your skills. my email is 150mikkimase@gmail.com and my cell is 6467097799. please contact me, Mikki, at any time suitable for you. i look forward to your inquiries and am excited to move forward with my new partner(s).","title":"Entrepreneur searching for developer counterpart","updated_at":"2023-09-07T02:51:02Z","url":""},{"_highlightResult":{"author":{"fullyHighlighted":true,"matchLevel":"partial","matchedWords":["thunder"],"value":"thund3rp"},"story_text":{"matchLevel":"none","matchedWords":[],"value":""},"title":{"matchLevel":"none","matchedWords":[],"value":"Al-Qaida comunications hacked.."},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["compute"],"value":"http://www.computerworld.com/s/article/9218050/Al_Qaida_communications_knocked_offline_says_expert_"}},"_tags":["story","author_thund3rp","story_2713224"],"author":"thund3rp","created_at":"2011-06-30T09:16:40Z","created_at_i":1309425400,"num_comments":0,"objectID":"2713224","points":2,"story_id":2713224,"story_text":"","title":"Al-Qaida comunications hacked..","updated_at":"2024-09-19T17:47:33Z","url":"http://www.computerworld.com/s/article/9218050/Al_Qaida_communications_knocked_offline_says_expert_"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"vishnukool"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["thunder","compute"],"value":"Hey HN, I\u2019m Vishnu from Agent37 (https://agent37.com). We build sandboxes for running AI agents at under 10% of the cost of most sandbox providers. Our customers spend an average of $0.37/month per sandbox.
You can view a video walkthrough of the product here: https://youtu.be/DK5ISR2tbNo
We started as one of those OpenClaw wrapper companies back in February (yup.. judge me). In time we added more and more harnesses.
One of the goals I had at the time was to offer a free plan to users, so as to gain more market share. But it was too cost prohibitive to do it right away. Especially since we were bootstrapped and servers were running on my savings.
Today we\u2019re actually able to offer the free plan thanks to the optimizations we\u2019ve made since. Over 90% of the new signups we have are just free plan users and they host their personal Hermes/Openclaw/Claude Code etc on our cloud. Our infrastructure cost for the free sandboxes averages around $0.05 to $0.10 per user per month.
Here are some techniques that have worked for lowering cost:
* When you look at agent harnesses, they have certain patterns in the way they consume CPU and RAM. They need CPU in bursts when they\u2019re initializing or performing certain tasks, but beyond that, they\u2019re often not peaking. They need more RAM for things like browser and computer-use tasks. Once the task is done, usage often comes back down. This is a key insight in lowering cost, so by efficiently managing the CPU/RAM utilization we could lower the cost.
* Secondly, you have to skip the hyperscalers when renting the Hosts. I worked at Amazon before and they have 80-90% margins on VMs. So it was an obvious choice once I looked at the prices to go one layer down to vendors like Hetzner and OVH. This alone was probably one of the biggest cost levers in reducing the cost.
* Thirdly, you also want to pick a technology stack that lets you efficiently cache and share resources across the host. We use gVisor for better utilization and fast boot times. It lets us use available host memory across sandboxes without reserving a full VM\u2019s memory for each one. There are tradeoffs: our gVisor sandboxes don\u2019t support workloads like Docker-in-Docker or Android emulators.
* And finally the size and shape of the hosts you buy also makes a big difference. RAM is typically the bottleneck when it comes to cost and not the CPU. This is specifically true for agent harness workloads, so when buying servers, we optimize for packing more RAM into a given host.
We also host full VM sandboxes for workloads that need them, including running Docker at a slightly higher price, but they are still pretty low cost given the rest of the techniques.
Overall, my learning has been that just by focusing on a niche of sandbox usecases, and optimizing for it you can give a much better cost/performance tradeoff.
I was previously a Senior Software Engineer at Amazon and an Engineering Manager at Quantcast. I built Amazon's internal AI chatbot, Diya, and their internal coding harness, Ziya.
If you\u2019re someone who\u2019s looking to run agents on sandbox and are exploring a sandbox vendor, come check us out. We\u2019re particularly looking for developers building messaging agents to try their workloads and challenge the limits and economics of our platform.
Thanks!
https://agent37.com"},"title":{"matchLevel":"none","matchedWords":[],"value":"Launch HN: Agent37 (YC F26) \u2013 Low-cost agent sandboxes for running agents"}},"_tags":["story","author_vishnukool","story_50059096","launch_hn"],"author":"vishnukool","children":[50059903,50059919,50060033,50060144,50060163,50060240,50060249,50060422,50060514,50060717,50061240,50061261,50062588,50063391,50063837,50063940,50064023,50064051,50064192,50064886,50065037,50065139,50065149,50065483,50065802,50065960,50066866,50066903,50066993,50067360,50067580,50067623,50067961,50068270,50068295,50068675,50068956,50069352,50070298,50070586,50070631,50070894,50071898,50073959,50076725,50083486,50085671,50089075,50101271,50111787],"created_at":"2026-10-12T16:46:34Z","created_at_i":1791823594,"num_comments":124,"objectID":"50059096","points":160,"story_id":50059096,"story_text":"Hey HN, I\u2019m Vishnu from Agent37 (https://agent37.com). We build sandboxes for running AI agents at under 10% of the cost of most sandbox providers. Our customers spend an average of $0.37/month per sandbox.
You can view a video walkthrough of the product here: https://youtu.be/DK5ISR2tbNo
We started as one of those OpenClaw wrapper companies back in February (yup.. judge me). In time we added more and more harnesses.
One of the goals I had at the time was to offer a free plan to users, so as to gain more market share. But it was too cost prohibitive to do it right away. Especially since we were bootstrapped and servers were running on my savings.
Today we\u2019re actually able to offer the free plan thanks to the optimizations we\u2019ve made since. Over 90% of the new signups we have are just free plan users and they host their personal Hermes/Openclaw/Claude Code etc on our cloud. Our infrastructure cost for the free sandboxes averages around $0.05 to $0.10 per user per month.
Here are some techniques that have worked for lowering cost:
* When you look at agent harnesses, they have certain patterns in the way they consume CPU and RAM. They need CPU in bursts when they\u2019re initializing or performing certain tasks, but beyond that, they\u2019re often not peaking. They need more RAM for things like browser and computer-use tasks. Once the task is done, usage often comes back down. This is a key insight in lowering cost, so by efficiently managing the CPU/RAM utilization we could lower the cost.
* Secondly, you have to skip the hyperscalers when renting the Hosts. I worked at Amazon before and they have 80-90% margins on VMs. So it was an obvious choice once I looked at the prices to go one layer down to vendors like Hetzner and OVH. This alone was probably one of the biggest cost levers in reducing the cost.
* Thirdly, you also want to pick a technology stack that lets you efficiently cache and share resources across the host. We use gVisor for better utilization and fast boot times. It lets us use available host memory across sandboxes without reserving a full VM\u2019s memory for each one. There are tradeoffs: our gVisor sandboxes don\u2019t support workloads like Docker-in-Docker or Android emulators.
* And finally the size and shape of the hosts you buy also makes a big difference. RAM is typically the bottleneck when it comes to cost and not the CPU. This is specifically true for agent harness workloads, so when buying servers, we optimize for packing more RAM into a given host.
We also host full VM sandboxes for workloads that need them, including running Docker at a slightly higher price, but they are still pretty low cost given the rest of the techniques.
Overall, my learning has been that just by focusing on a niche of sandbox usecases, and optimizing for it you can give a much better cost/performance tradeoff.
I was previously a Senior Software Engineer at Amazon and an Engineering Manager at Quantcast. I built Amazon's internal AI chatbot, Diya, and their internal coding harness, Ziya.
If you\u2019re someone who\u2019s looking to run agents on sandbox and are exploring a sandbox vendor, come check us out. We\u2019re particularly looking for developers building messaging agents to try their workloads and challenge the limits and economics of our platform.
Thanks!
https://agent37.com","title":"Launch HN: Agent37 (YC F26) \u2013 Low-cost agent sandboxes for running agents","updated_at":"2026-10-16T05:26:33Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"SaarthRanka"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["thunder","compute"],"value":"Hi HN! I'm Saarth. I'm in 9th grade, and I built Atticus.
Atticus is a fast and cheap Mac assistant that can do tasks for you. You hold down Control-Option-Space, speak or type your task, be it complex or a simple browser search, and Atticus will get on it.
The interesting part is that simple tasks, like opening apps, playing a Spotify playlist, or web queries, don't touch AI, and instead are processed locally and typically finish in under a second. For example, prompting Atticus to open apps takes around 0.08 seconds, sometimes even finishing before you finish your sentence, sort of like Jev.
Any tasks that are more complicated will be sent to Claude via claude -p, meaning you never need an API key and Atticus will run on your existing Claude Pro or Max plan. By using this approach, costs are also cut down immensely, with, on average, tasks costing a fraction of a cent.
To achieve Atticus' speed, it remembers shortcuts from each task it completes. When it does a task right for the first time, it saves its steps and replays them the next time without AI. For example, a workflow of creating a new meeting took about 30 seconds the first time around. Atticus then logs the exact steps to memory and takes 5-8 seconds on the next pass. To add to this, Atticus also learns about you as you use it more and from what's on your computer, including your name, your apps, and things you want it to remember, personalizing its answers to you.
To combat the safety concerns over giving an AI unrestricted access to your device, Atticus has toggleable restrictions, so before anything is deleted, sent, bought, or written to a file, the user is prompted before that action.
Some features I'm currently working on include a meeting copilot that capitalizes on Atticus' speed, allowing Atticus to listen in on meetings and answer questions in real time as they come in, personalized to the context of the meeting and what it knows about you.
To run Atticus, macOS 26 or higher is required for its voice transcription features. The voice transcription is also done locally by macOS.
Atticus is free, open source, and also very fresh, and I'm completely aware of its many rough edges, which is why I'd love to hear from everyone about what breaks and what you guys would like to see Atticus do!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Atticus \u2013 Hold-to-talk voice assistant for macOS that runs on Claude -p"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://github.com/SaarthurR/atticus-mac"}},"_tags":["story","author_SaarthRanka","story_50049588","show_hn"],"author":"SaarthRanka","children":[50049589,50052190,50052949,50053168,50053903,50054667,50055941,50058964,50064434,50064732,50065467,50065503,50065511,50066597,50067423,50067562,50068414,50071941,50080444,50080477],"created_at":"2026-10-12T00:15:12Z","created_at_i":1791764112,"num_comments":30,"objectID":"50049588","points":115,"story_id":50049588,"story_text":"Hi HN! I'm Saarth. I'm in 9th grade, and I built Atticus.
Atticus is a fast and cheap Mac assistant that can do tasks for you. You hold down Control-Option-Space, speak or type your task, be it complex or a simple browser search, and Atticus will get on it.
The interesting part is that simple tasks, like opening apps, playing a Spotify playlist, or web queries, don't touch AI, and instead are processed locally and typically finish in under a second. For example, prompting Atticus to open apps takes around 0.08 seconds, sometimes even finishing before you finish your sentence, sort of like Jev.
Any tasks that are more complicated will be sent to Claude via claude -p, meaning you never need an API key and Atticus will run on your existing Claude Pro or Max plan. By using this approach, costs are also cut down immensely, with, on average, tasks costing a fraction of a cent.
To achieve Atticus' speed, it remembers shortcuts from each task it completes. When it does a task right for the first time, it saves its steps and replays them the next time without AI. For example, a workflow of creating a new meeting took about 30 seconds the first time around. Atticus then logs the exact steps to memory and takes 5-8 seconds on the next pass. To add to this, Atticus also learns about you as you use it more and from what's on your computer, including your name, your apps, and things you want it to remember, personalizing its answers to you.
To combat the safety concerns over giving an AI unrestricted access to your device, Atticus has toggleable restrictions, so before anything is deleted, sent, bought, or written to a file, the user is prompted before that action.
Some features I'm currently working on include a meeting copilot that capitalizes on Atticus' speed, allowing Atticus to listen in on meetings and answer questions in real time as they come in, personalized to the context of the meeting and what it knows about you.
To run Atticus, macOS 26 or higher is required for its voice transcription features. The voice transcription is also done locally by macOS.
Atticus is free, open source, and also very fresh, and I'm completely aware of its many rough edges, which is why I'd love to hear from everyone about what breaks and what you guys would like to see Atticus do!","title":"Show HN: Atticus \u2013 Hold-to-talk voice assistant for macOS that runs on Claude -p","updated_at":"2026-10-16T04:29:48Z","url":"https://github.com/SaarthurR/atticus-mac"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"thedracle"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["thunder","compute"],"value":"Hi HN!
I\u2019m Jason Thomas, one of the founders of CoScreen.
As some of you may know, Datadog sunset the CoScreen service on July 31st.
Four of us from the original CoScreen team have spent the last couple months building a free and open source pair programming tool inspired by those that came before us: It\u2019s called Petal (https://github.com/kiruna-labs/petal)
Some history first, because when we started CoScreen we had hoped to break the sad and established pattern of this industry.
Screenhero was an exceptional tool that a lot of us loved for pairing. Multiple cursors, low latency, you could click your partner\u2019s editor, and just start typing. It felt like magic while we had it. Then Slack bought it in 2015, and shut down the standalone app in December 2017.
When Till, Max, and I started CoScreen, we were aware of this history, and wanted to build an even more magical collaboration and screen sharing experience: the entire concept was \u201cthrow a window from your laptop up on someone else\u2019s screen, and start collaborating with zero friction.\u201d
Datadog acquired us in 2022, and thus began the great adventure of adapting our real-time technology into Datadog products, while also continue developing CoScreen as a standalone tool from the Datadog ecosystem. Datadog invested in the product for years and we\u2019re thankful for that. But CoScreen itself proved, like so many other screen sharing tool acquisitions, to be too tangential from the customer base and core mission. We instead built other communication features in Datadog, the integrations with other video platforms, and also its wildly successful MCP server.\nWe weren't the only ones. Around was bought by Miro and shut down in March 2025. Multi (formerly Remotion) was acqui-hired by OpenAI in 2024 and shut down about a month later. It's nearly always the same story: a small team builds something hard that people love, a platform buys it for the team or the feature, and the standalone product doesn't survive the integration.\nSo we decided firmly to do things differently this time. Petal is completely free and open source and we aren't monetizing it. If we lose interest or get hit by a bus, you can self-host it or fork it, and nobody can sunset it on you. Honestly it's a love letter to everyone in this community that supported us, believed in us, and to liberate collaborative screen sharing from this vicious cycle.
We hope through this we can help foster a large and vibrant community around innovating in this product sector.\nWhat Petal does: shared windows show up as real windows on your desktop, movable and resizable, not tiles in a grid. Several people can share at once, everyone gets a labeled cursor, and you can click into any shared window to type and scroll. Rooms are persistent, so you click once to join "eng-sync" instead of passing links around.\nIt runs on macOS and Windows (Linux is next, this time for real and you can help us!), plus a web app. Media goes over LiveKit. You can use our hosted instance with zero config, or point the app at your own LiveKit server with one docker-compose if you want the traffic to stay on your network or even use an entirely different video backend if you want to.\nThe part that still feels strange to us: CoScreen took almost six years to build. The first version of Petal took under eight weeks, from scratch, with frontier models like Fable and Astra doing a lot of the work. Real-time video and window capture is some of the most unforgiving code we have ever written, so we didn't expect that. We will do a proper writeup on what worked and what didn't, and are happy to answer questions about it here.\nTill, Darren, Max and I will be in the thread."},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Petal, open-source CoScreen/ScreenHero finally free"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://petal.live/"}},"_tags":["story","author_thedracle","story_50092470","show_hn"],"author":"thedracle","children":[50092660,50093031,50093211,50093394,50094088,50094634,50094664,50095268,50095294,50095807,50095889,50095983,50096872,50097516,50098031,50098436,50098858,50099660,50099892,50101209,50101452,50101648],"created_at":"2026-10-14T18:23:14Z","created_at_i":1792002194,"num_comments":41,"objectID":"50092470","points":215,"story_id":50092470,"story_text":"Hi HN!
I\u2019m Jason Thomas, one of the founders of CoScreen.
As some of you may know, Datadog sunset the CoScreen service on July 31st.
Four of us from the original CoScreen team have spent the last couple months building a free and open source pair programming tool inspired by those that came before us: It\u2019s called Petal (https://github.com/kiruna-labs/petal)
Some history first, because when we started CoScreen we had hoped to break the sad and established pattern of this industry.
Screenhero was an exceptional tool that a lot of us loved for pairing. Multiple cursors, low latency, you could click your partner\u2019s editor, and just start typing. It felt like magic while we had it. Then Slack bought it in 2015, and shut down the standalone app in December 2017.
When Till, Max, and I started CoScreen, we were aware of this history, and wanted to build an even more magical collaboration and screen sharing experience: the entire concept was \u201cthrow a window from your laptop up on someone else\u2019s screen, and start collaborating with zero friction.\u201d
Datadog acquired us in 2022, and thus began the great adventure of adapting our real-time technology into Datadog products, while also continue developing CoScreen as a standalone tool from the Datadog ecosystem. Datadog invested in the product for years and we\u2019re thankful for that. But CoScreen itself proved, like so many other screen sharing tool acquisitions, to be too tangential from the customer base and core mission. We instead built other communication features in Datadog, the integrations with other video platforms, and also its wildly successful MCP server.\nWe weren't the only ones. Around was bought by Miro and shut down in March 2025. Multi (formerly Remotion) was acqui-hired by OpenAI in 2024 and shut down about a month later. It's nearly always the same story: a small team builds something hard that people love, a platform buys it for the team or the feature, and the standalone product doesn't survive the integration.\nSo we decided firmly to do things differently this time. Petal is completely free and open source and we aren't monetizing it. If we lose interest or get hit by a bus, you can self-host it or fork it, and nobody can sunset it on you. Honestly it's a love letter to everyone in this community that supported us, believed in us, and to liberate collaborative screen sharing from this vicious cycle.
We hope through this we can help foster a large and vibrant community around innovating in this product sector.\nWhat Petal does: shared windows show up as real windows on your desktop, movable and resizable, not tiles in a grid. Several people can share at once, everyone gets a labeled cursor, and you can click into any shared window to type and scroll. Rooms are persistent, so you click once to join "eng-sync" instead of passing links around.\nIt runs on macOS and Windows (Linux is next, this time for real and you can help us!), plus a web app. Media goes over LiveKit. You can use our hosted instance with zero config, or point the app at your own LiveKit server with one docker-compose if you want the traffic to stay on your network or even use an entirely different video backend if you want to.\nThe part that still feels strange to us: CoScreen took almost six years to build. The first version of Petal took under eight weeks, from scratch, with frontier models like Fable and Astra doing a lot of the work. Real-time video and window capture is some of the most unforgiving code we have ever written, so we didn't expect that. We will do a proper writeup on what worked and what didn't, and are happy to answer questions about it here.\nTill, Darren, Max and I will be in the thread.","title":"Show HN: Petal, open-source CoScreen/ScreenHero finally free","updated_at":"2026-10-16T05:31:19Z","url":"https://petal.live/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ms209495"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["thunder","compute"],"value":"I want to show you the game I created recently and tell you whole the story/experience - including backend/latency/scalability.
You can try it quickly without an account - I created game demo rooms (EU/US/JP) - links below - just be aware that anyone with a link can access join the game.
Let's start with short description of the game itself:
This is the TowerPanic game - you are standing on one of four towers, during each round some of towers crack, and some of these lean, and finally one of these colapse - you need to watch out to not be on the failing one.\nIt's a multiplayer game for whole family.\nThis can be also used as an icebreaking game - for people in any age.\nThe game is safe for children - no violence, no strangers, no chat, no ads.\nParent creates an account (free 14-day trial, then a paid pass) and starts the game - others join using a QR code or a link.
Behind the game there is an interesting backend engine - I willdescribe it in a moment.
Let's start with the history of this whole project:\n- I was playing with asyncio and fastapi\n- As an exercise I created a reliable chat application (multiple rooms, in each room each user can reliably broadcast a message to all other users)\n- I extracted communication to separate layer - creating network/application layer interfaces - 'reliable client' interface for client and 'state machine' interface for the service on server.\n- With these interfaces - I was able to create other service - an 'sgame' (simple game). With verified interfaces and created services (chat, sgame) I was able to ask AI to generate other tools/games - I was impressed by these results\n- the games/tools worked well - I deployed them on a remote server - I was surprised that the latency was very good - almost everything worked smoothly
I tested the tools and games with my children and one of the game was really great.\nI decided to make this game public - I chose 'TowerPanic' name for it.
I prepared an account/payment server for this game (as a separate FastAPI application).\nThis was very good choice:\n- the game servers keep rooms only in memory - nothing persistent (rooms have expiration date)\n- the account server keeps all persistent data (it uses database)
Because of this split I was able to figure out how to upgrade and scale frontend and backend independently - without significantly impacting users.
I optimized the communication layer - so that it works in finite memory - only finite number of recent messages is kept, and I sent a full service snapshot (full state) to the lagging clients.
The backend process (FastAPI/python) is limited by python GIL (only one core can be utilized) - to utilize all server cores I used HAProxy with consistent hashing to distribute rooms - so that users can be redirected to proper backend process.
I tested how many rooms a single server can handle - it handles hundreds of active rooms in parallel with no significant latency drop.
Before making it public I noticed that latency in other region of world might be higher and this could influence UX - I confirmed this using artificial delay on my remote server.
I decided to deploy the application in three regions - EU, US, JP - to ensure low latency (below 150ms) in most of the world.\nThe room is created in region closest to the user (best latency from the browser to backend).
I needed also to prepare some legal documents (terms, privacy, refunds) - thanks to AI this was much easier.\nI'm a backend engineer, frontend/UI/UX were designed by me but code was generated by AI.
To summerize all this work:\n- I'm happy that I was able to deliver complete game/complete product\n- I think the whole work took me about 20 evenings - I'm impressed that it was so short - I think this is because the technology I used is popular and all fits well \n- AI was great as a peer (to discuss the options for distributing the load, discuss were to locate servers, or discuss how auth/payments work under the hood, ... - based on this I could choose the best option)\n- AI was great for doing the non work which I'm doing ineffectively (frontend/UI/legal docs).
The game is available at https://towerpanic.com/?ref=hn
I will appreciate your comments/feedback:
Product:\n - did you meet any obstacles to play ?\n - how smoothly it worked ? - please specify your location if it didn't work well \n - most important thing/things to improve ?\n - anything else ...\nSolution/Engineering:\n - one thing you will change in the architecture / approach ?\n - anything else ...
If you don't want to register you can play directly using below links (2 rooms for each location, max 10 players in each room/board). Choose best location:
DEMO ROOMS (expire: 19 October 2026 at 22:00 UTC):
EU:\nhttps://engine.prod-de-1.towerpanic.com/b/hn1/join/AUB3lS3Y_...\nhttps://engine.prod-de-1.towerpanic.com/b/hn4/join/ASlA3J43U...
US:\nhttps://engine.prod-us-1.towerpanic.com/b/hn2/join/AUCCS_x9d...\nhttps://engine.prod-us-1.towerpanic.com/b/hn5/join/AVQbs_01N...
JP:\nhttps://engine.prod-jp-1.towerpanic.com/b/hn3/join/Afj600qp5...\nhttps://engine.prod-jp-1.towerpanic.com/b/hn6/join/AUpDUuB1W..."},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: TowerPanic \u2013 a family party game on a small FastAPI realtime engine"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://towerpanic.com/"}},"_tags":["story","author_ms209495","story_50098245","show_hn"],"author":"ms209495","children":[50098311,50103395],"created_at":"2026-10-15T02:02:26Z","created_at_i":1792029746,"num_comments":5,"objectID":"50098245","points":3,"story_id":50098245,"story_text":"I want to show you the game I created recently and tell you whole the story/experience - including backend/latency/scalability.
You can try it quickly without an account - I created game demo rooms (EU/US/JP) - links below - just be aware that anyone with a link can access join the game.
Let's start with short description of the game itself:
This is the TowerPanic game - you are standing on one of four towers, during each round some of towers crack, and some of these lean, and finally one of these colapse - you need to watch out to not be on the failing one.\nIt's a multiplayer game for whole family.\nThis can be also used as an icebreaking game - for people in any age.\nThe game is safe for children - no violence, no strangers, no chat, no ads.\nParent creates an account (free 14-day trial, then a paid pass) and starts the game - others join using a QR code or a link.
Behind the game there is an interesting backend engine - I willdescribe it in a moment.
Let's start with the history of this whole project:\n- I was playing with asyncio and fastapi\n- As an exercise I created a reliable chat application (multiple rooms, in each room each user can reliably broadcast a message to all other users)\n- I extracted communication to separate layer - creating network/application layer interfaces - 'reliable client' interface for client and 'state machine' interface for the service on server.\n- With these interfaces - I was able to create other service - an 'sgame' (simple game). With verified interfaces and created services (chat, sgame) I was able to ask AI to generate other tools/games - I was impressed by these results\n- the games/tools worked well - I deployed them on a remote server - I was surprised that the latency was very good - almost everything worked smoothly
I tested the tools and games with my children and one of the game was really great.\nI decided to make this game public - I chose 'TowerPanic' name for it.
I prepared an account/payment server for this game (as a separate FastAPI application).\nThis was very good choice:\n- the game servers keep rooms only in memory - nothing persistent (rooms have expiration date)\n- the account server keeps all persistent data (it uses database)
Because of this split I was able to figure out how to upgrade and scale frontend and backend independently - without significantly impacting users.
I optimized the communication layer - so that it works in finite memory - only finite number of recent messages is kept, and I sent a full service snapshot (full state) to the lagging clients.
The backend process (FastAPI/python) is limited by python GIL (only one core can be utilized) - to utilize all server cores I used HAProxy with consistent hashing to distribute rooms - so that users can be redirected to proper backend process.
I tested how many rooms a single server can handle - it handles hundreds of active rooms in parallel with no significant latency drop.
Before making it public I noticed that latency in other region of world might be higher and this could influence UX - I confirmed this using artificial delay on my remote server.
I decided to deploy the application in three regions - EU, US, JP - to ensure low latency (below 150ms) in most of the world.\nThe room is created in region closest to the user (best latency from the browser to backend).
I needed also to prepare some legal documents (terms, privacy, refunds) - thanks to AI this was much easier.\nI'm a backend engineer, frontend/UI/UX were designed by me but code was generated by AI.
To summerize all this work:\n- I'm happy that I was able to deliver complete game/complete product\n- I think the whole work took me about 20 evenings - I'm impressed that it was so short - I think this is because the technology I used is popular and all fits well \n- AI was great as a peer (to discuss the options for distributing the load, discuss were to locate servers, or discuss how auth/payments work under the hood, ... - based on this I could choose the best option)\n- AI was great for doing the non work which I'm doing ineffectively (frontend/UI/legal docs).
The game is available at https://towerpanic.com/?ref=hn
I will appreciate your comments/feedback:
Product:\n - did you meet any obstacles to play ?\n - how smoothly it worked ? - please specify your location if it didn't work well \n - most important thing/things to improve ?\n - anything else ...\nSolution/Engineering:\n - one thing you will change in the architecture / approach ?\n - anything else ...
If you don't want to register you can play directly using below links (2 rooms for each location, max 10 players in each room/board). Choose best location:
DEMO ROOMS (expire: 19 October 2026 at 22:00 UTC):
EU:\nhttps://engine.prod-de-1.towerpanic.com/b/hn1/join/AUB3lS3Y_...\nhttps://engine.prod-de-1.towerpanic.com/b/hn4/join/ASlA3J43U...
US:\nhttps://engine.prod-us-1.towerpanic.com/b/hn2/join/AUCCS_x9d...\nhttps://engine.prod-us-1.towerpanic.com/b/hn5/join/AVQbs_01N...
JP:\nhttps://engine.prod-jp-1.towerpanic.com/b/hn3/join/Afj600qp5...\nhttps://engine.prod-jp-1.towerpanic.com/b/hn6/join/AUpDUuB1W...","title":"Show HN: TowerPanic \u2013 a family party game on a small FastAPI realtime engine","updated_at":"2026-10-16T03:33:19Z","url":"https://towerpanic.com/"}],"hitsPerPage":50,"nbHits":14,"nbPages":1,"page":0,"params":"query=Thunder+Compute&tags=story&hitsPerPage=50&advancedSyntax=true&analyticsTags=backend","processingTimeMS":21,"processingTimingsMS":{"_request":{"queue":118,"roundTrip":21},"afterFetch":{"format":{"highlighting":2,"total":3}},"fetch":{"query":17,"scanning":2,"total":20},"total":21},"query":"Thunder Compute","serverTimeMS":142}