What\u2019s missing from the current conversation is the question:\n\u201cWhat comes after LEO?\u201d
Because the answer determines whether we build something scalable or something we\u2019ll have to replace later at enormous cost.
1. LEO is useful as a starting point \u2014 but fundamentally limited. \nLEO offers fast iteration and easy access, but it also brings inherent constraints:\n\u2013 orbital congestion\n\u2013 debris accumulation\n\u2013 interference with astronomy\n\u2013 long\u2011term shadowing concerns if scaled\n\u2013 limited energy density
These aren\u2019t hypothetical future problems.\nThey\u2019re structural properties of LEO.
2. Solar\u2011proximal orbits (Venus/Mercury) are the physically optimal destination. \nSolar flux increases dramatically as you move inward:\n\u2013 Venus orbit: ~2\u00d7 Earth\n\u2013 Mercury orbit: ~11\u00d7 Earth
Even for early deployments, this means:\n\u2013 more energy per square meter\n\u2013 fewer satellites for the same output\n\u2013 lower mass per watt\n\u2013 simpler thermal management\n\u2013 better long\u2011term economics
You don\u2019t need to wait decades for these advantages to matter.\nThey\u2019re true from day one.
3. The Moon is a better long\u2011term platform than LEO for maintenance and scaling. \nIf we expect lunar manufacturing and launch infrastructure to grow, then the Moon is also a natural place to host compute:\n\u2013 stable surface\n\u2013 easy robotic maintenance\n\u2013 no Kessler risk\n\u2013 nuclear reactors allowed\n\u2013 no interference with Earth\u2019s sky\n\u2013 no cumulative shadowing on Earth\u2019s climate
The Moon is a scalable compute platform in a way LEO simply isn\u2019t.
4. LEO as a trampoline is fine \u2014 but only if we already know the direction of the jump. \nThe real risk is not starting in LEO.\nThe risk is staying in LEO by inertia, simply because it\u2019s familiar.
Humanity tends to fix architectures only after they break.\nBut in this case, waiting until LEO is saturated will make the transition vastly more expensive.
We should be discussing the post\u2011LEO architecture now, not when the bottlenecks appear.
5. A civilization aiming at Kardashev I cannot rely on Earth\u2011adjacent infrastructure. \nA Type I trajectory naturally pushes us toward:\n\u2013 lunar compute arrays\n\u2013 Venus/Mercury solar compute rings\n\u2013 heliocentric data infrastructure\n\u2013 Dyson\u2011compatible architectures
These are not sci\u2011fi endpoints.\nThey\u2019re the physically optimal places to put large\u2011scale compute and energy harvesting.
Conclusion: \nLEO may be a practical starting point, but it is not the right long\u2011term architecture.\nIf we want orbital compute that is scalable, safe, and economically efficient, we need to start planning beyond LEO now \u2014 toward solar\u2011proximal orbits and lunar platforms where energy is abundant and orbital congestion is irrelevant.
The conversation about \u201cwhat comes after LEO\u201d is overdue.\nIt\u2019s time to start it."},"title":{"matchLevel":"none","matchedWords":[],"value":"We Need to Start Planning Beyond Leo"}},"_tags":["story","author_NahumMartinez","story_48569203","ask_hn"],"author":"NahumMartinez","created_at":"2026-06-17T12:04:17Z","created_at_i":1781697857,"num_comments":0,"objectID":"48569203","points":7,"story_id":48569203,"story_text":"Orbital compute farms are coming. Musk\u2019s TERAFAB vision is an important early step, and LEO may be a practical place to begin.\nBut if we think even slightly beyond the immediate engineering horizon, it becomes clear that LEO cannot be the long\u2011term architecture for civilization\u2011scale compute.
What\u2019s missing from the current conversation is the question:\n\u201cWhat comes after LEO?\u201d
Because the answer determines whether we build something scalable or something we\u2019ll have to replace later at enormous cost.
1. LEO is useful as a starting point \u2014 but fundamentally limited. \nLEO offers fast iteration and easy access, but it also brings inherent constraints:\n\u2013 orbital congestion\n\u2013 debris accumulation\n\u2013 interference with astronomy\n\u2013 long\u2011term shadowing concerns if scaled\n\u2013 limited energy density
These aren\u2019t hypothetical future problems.\nThey\u2019re structural properties of LEO.
2. Solar\u2011proximal orbits (Venus/Mercury) are the physically optimal destination. \nSolar flux increases dramatically as you move inward:\n\u2013 Venus orbit: ~2\u00d7 Earth\n\u2013 Mercury orbit: ~11\u00d7 Earth
Even for early deployments, this means:\n\u2013 more energy per square meter\n\u2013 fewer satellites for the same output\n\u2013 lower mass per watt\n\u2013 simpler thermal management\n\u2013 better long\u2011term economics
You don\u2019t need to wait decades for these advantages to matter.\nThey\u2019re true from day one.
3. The Moon is a better long\u2011term platform than LEO for maintenance and scaling. \nIf we expect lunar manufacturing and launch infrastructure to grow, then the Moon is also a natural place to host compute:\n\u2013 stable surface\n\u2013 easy robotic maintenance\n\u2013 no Kessler risk\n\u2013 nuclear reactors allowed\n\u2013 no interference with Earth\u2019s sky\n\u2013 no cumulative shadowing on Earth\u2019s climate
The Moon is a scalable compute platform in a way LEO simply isn\u2019t.
4. LEO as a trampoline is fine \u2014 but only if we already know the direction of the jump. \nThe real risk is not starting in LEO.\nThe risk is staying in LEO by inertia, simply because it\u2019s familiar.
Humanity tends to fix architectures only after they break.\nBut in this case, waiting until LEO is saturated will make the transition vastly more expensive.
We should be discussing the post\u2011LEO architecture now, not when the bottlenecks appear.
5. A civilization aiming at Kardashev I cannot rely on Earth\u2011adjacent infrastructure. \nA Type I trajectory naturally pushes us toward:\n\u2013 lunar compute arrays\n\u2013 Venus/Mercury solar compute rings\n\u2013 heliocentric data infrastructure\n\u2013 Dyson\u2011compatible architectures
These are not sci\u2011fi endpoints.\nThey\u2019re the physically optimal places to put large\u2011scale compute and energy harvesting.
Conclusion: \nLEO may be a practical starting point, but it is not the right long\u2011term architecture.\nIf we want orbital compute that is scalable, safe, and economically efficient, we need to start planning beyond LEO now \u2014 toward solar\u2011proximal orbits and lunar platforms where energy is abundant and orbital congestion is irrelevant.
The conversation about \u201cwhat comes after LEO\u201d is overdue.\nIt\u2019s time to start it.","title":"We Need to Start Planning Beyond Leo","updated_at":"2026-06-17T19:59:13Z"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"fiatjaf"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["lunar","outpost"],"value":"Show HN: Test { fuzzyset, sifter, lunr, fuse, levi } output at the same time"},"url":{"matchLevel":"none","matchedWords":[],"value":"http://fiatjaf.alhur.es/js-search-engines-comparison/"}},"_tags":["story","author_fiatjaf","story_12104953","show_hn"],"author":"fiatjaf","children":[12104955],"created_at":"2016-07-16T02:22:03Z","created_at_i":1468635723,"num_comments":1,"objectID":"12104953","points":1,"story_id":12104953,"title":"Show HN: Test { fuzzyset, sifter, lunr, fuse, levi } output at the same time","updated_at":"2024-09-19T23:24:56Z","url":"http://fiatjaf.alhur.es/js-search-engines-comparison/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"bartdegoede"},"title":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["lunar","outpost"],"value":"Use Hugo Output Formats to generate Lunr index files for your static site search"},"url":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["lunar","outpost"],"value":"https://bart.degoe.de/use-hugo-output-formats-to-generate-lunr-index-files/"}},"_tags":["story","author_bartdegoede","story_20425443"],"author":"bartdegoede","children":[20426463],"created_at":"2019-07-12T23:29:30Z","created_at_i":1562974170,"num_comments":1,"objectID":"20425443","points":12,"story_id":20425443,"title":"Use Hugo Output Formats to generate Lunr index files for your static site search","updated_at":"2024-09-20T04:34:22Z","url":"https://bart.degoe.de/use-hugo-output-formats-to-generate-lunr-index-files/"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"nonatofabio"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["lunar","outpost"],"value":"Hey HN. I evaluated three agent frameworks for a homelab project, one had 400K lines of code (and 42K exposed instances on Shodan), one was 9 days old, and the third was so thin I'd rebuild most of it anyway. None worked for me.
So I built my own in ~2300 lines of Python. No frameworks, 8 runtime dependencies, 106 tests.
What it does:\n- Persistent memory via SQLite (FTS5 keyword search + sqlite-vec embeddings + recency decay, fused with Reciprocal Rank Fusion)\n- MCP tool integration \u2014 add capabilities by editing a JSON file\n- Native tools with safety guardrails (bash blocklist, timeouts, output caps)\n- Discord interface with session isolation\n- Structured JSON logging for every operation\n- Conversation compression for effectively infinite context
Runs locally on 2x RTX 3090 with Qwen3-Coder-Next via llama-server. No cloud APIs.
The design philosophy was: don't build what you don't need, but don't block the insertion points. For example, my AI firewall isn't built yet, but all LLM traffic goes through a single configurable URL, swapping in a filtering proxy is a config change I'll do later.
DESIGN.md documents the reasoning behind every architectural decision. Tests mock the LLM client so you can run them on a laptop.
GitHub: https://github.com/nonatofabio/luna-agent\nBlog post with full technical deep-dive: https://nonatofabio.github.io/blog/post.html?slug=luna_agent
Happy to answer questions about any of the design tradeoffs."},"title":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["lunar"],"value":"Show HN: Luna Agent \u2013 Custom AI agent in ~2300 lines of Python, no frameworks"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["lunar"],"value":"https://nonatofabio.github.io/blog/post.html?slug=luna_agent"}},"_tags":["story","author_nonatofabio","story_47291343","show_hn"],"author":"nonatofabio","created_at":"2026-03-07T20:49:53Z","created_at_i":1772916593,"num_comments":0,"objectID":"47291343","points":1,"story_id":47291343,"story_text":"Hey HN. I evaluated three agent frameworks for a homelab project, one had 400K lines of code (and 42K exposed instances on Shodan), one was 9 days old, and the third was so thin I'd rebuild most of it anyway. None worked for me.
So I built my own in ~2300 lines of Python. No frameworks, 8 runtime dependencies, 106 tests.
What it does:\n- Persistent memory via SQLite (FTS5 keyword search + sqlite-vec embeddings + recency decay, fused with Reciprocal Rank Fusion)\n- MCP tool integration \u2014 add capabilities by editing a JSON file\n- Native tools with safety guardrails (bash blocklist, timeouts, output caps)\n- Discord interface with session isolation\n- Structured JSON logging for every operation\n- Conversation compression for effectively infinite context
Runs locally on 2x RTX 3090 with Qwen3-Coder-Next via llama-server. No cloud APIs.
The design philosophy was: don't build what you don't need, but don't block the insertion points. For example, my AI firewall isn't built yet, but all LLM traffic goes through a single configurable URL, swapping in a filtering proxy is a config change I'll do later.
DESIGN.md documents the reasoning behind every architectural decision. Tests mock the LLM client so you can run them on a laptop.
GitHub: https://github.com/nonatofabio/luna-agent\nBlog post with full technical deep-dive: https://nonatofabio.github.io/blog/post.html?slug=luna_agent
Happy to answer questions about any of the design tradeoffs.","title":"Show HN: Luna Agent \u2013 Custom AI agent in ~2300 lines of Python, no frameworks","updated_at":"2026-03-07T20:54:04Z","url":"https://nonatofabio.github.io/blog/post.html?slug=luna_agent"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"ozankabak"},"story_text":{"fullyHighlighted":false,"matchLevel":"full","matchedWords":["lunar","outpost"],"value":"Hi HN! We are looking to gather some feedback on our serverless agent harness. The admittedly not-so-specific use case is to accelerate agent development and deployment. After building several custom/special-purpose agents for a few customers, we built this to accelerate our workflow at first, and now we are trying to understand whether it could be useful to others.
Our driver use case was development of specialist agents with a request/response lifecycle. Think of agents that have a well established input/output contract where they are expected to produce high-quality output (artifacts, responses etc.). Especially when the problem is in some verifiable domain and the LLM can iteratively refine a result to a final value that satisfies constraints or optimizes some goal.
The product is a coding agent skill + a serverless execution runtime with a harness that takes in a system prompt + Python functions as tools. The coding agent takes in the requirements from the user, and tries agent variants by executing prompt/tool variants it creates.
It works best for cases where you can think of how you can evaluate a candidate agent - when you describe this information to your coding agent, it often does a decent job at building candidate prompts, tools and even benchmarks.
Prompts and Python tools that the coding agent creates integrate with a harness that implements an FSM that is tuned to drive an iterative refinement process for verifiable domains. This tuning enables one to use small models like Luna to produce high quality results while keeping costs at a manageable level.
Our website is not 100% complete yet (some examples are missing write-ups, not all use cases we tried are there etc.), but the system is operational and docs are there.
Thanks!"},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Recurse \u2013 Develop and deploy specialist agents faster"},"url":{"matchLevel":"none","matchedWords":[],"value":"https://recurse.run"}},"_tags":["story","author_ozankabak","story_49850553","show_hn"],"author":"ozankabak","children":[49850554,49852344,49852631,49853075,49854563,49855584,49858661],"created_at":"2026-09-25T22:06:03Z","created_at_i":1790373963,"num_comments":4,"objectID":"49850553","points":5,"story_id":49850553,"story_text":"Hi HN! We are looking to gather some feedback on our serverless agent harness. The admittedly not-so-specific use case is to accelerate agent development and deployment. After building several custom/special-purpose agents for a few customers, we built this to accelerate our workflow at first, and now we are trying to understand whether it could be useful to others.
Our driver use case was development of specialist agents with a request/response lifecycle. Think of agents that have a well established input/output contract where they are expected to produce high-quality output (artifacts, responses etc.). Especially when the problem is in some verifiable domain and the LLM can iteratively refine a result to a final value that satisfies constraints or optimizes some goal.
The product is a coding agent skill + a serverless execution runtime with a harness that takes in a system prompt + Python functions as tools. The coding agent takes in the requirements from the user, and tries agent variants by executing prompt/tool variants it creates.
It works best for cases where you can think of how you can evaluate a candidate agent - when you describe this information to your coding agent, it often does a decent job at building candidate prompts, tools and even benchmarks.
Prompts and Python tools that the coding agent creates integrate with a harness that implements an FSM that is tuned to drive an iterative refinement process for verifiable domains. This tuning enables one to use small models like Luna to produce high quality results while keeping costs at a manageable level.
Our website is not 100% complete yet (some examples are missing write-ups, not all use cases we tried are there etc.), but the system is operational and docs are there.
Thanks!","title":"Show HN: Recurse \u2013 Develop and deploy specialist agents faster","updated_at":"2026-09-26T17:26:29Z","url":"https://recurse.run"},{"_highlightResult":{"author":{"matchLevel":"none","matchedWords":[],"value":"vincelt"},"story_text":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["outpost"],"value":"Hi HN,
I made a TypeScript library that unifies queries with different LLM providers like Anthropic and Gemini into the OpenAI format.
When playing around with new models like the new Gemini 2.0, it's frustrating to adapt to the differences in things like streaming, tool calling, json formatting, output formats, etc.
This library aims to make it easy to switch between providers, while keeping the OpenAI format, which seems to have become the de-facto standard. \nMaybe this can be helpful to others."},"title":{"matchLevel":"none","matchedWords":[],"value":"Show HN: Abso \u2013 TypeScript LLM Client"},"url":{"fullyHighlighted":false,"matchLevel":"partial","matchedWords":["lunar"],"value":"https://github.com/lunary-ai/abso"}},"_tags":["story","author_vincelt","story_43025829","show_hn"],"author":"vincelt","created_at":"2025-02-12T14:40:08Z","created_at_i":1739371208,"num_comments":0,"objectID":"43025829","points":1,"story_id":43025829,"story_text":"Hi HN,
I made a TypeScript library that unifies queries with different LLM providers like Anthropic and Gemini into the OpenAI format.
When playing around with new models like the new Gemini 2.0, it's frustrating to adapt to the differences in things like streaming, tool calling, json formatting, output formats, etc.
This library aims to make it easy to switch between providers, while keeping the OpenAI format, which seems to have become the de-facto standard. \nMaybe this can be helpful to others.","title":"Show HN: Abso \u2013 TypeScript LLM Client","updated_at":"2025-02-12T14:42:40Z","url":"https://github.com/lunary-ai/abso"}],"hitsPerPage":50,"nbHits":23,"nbPages":1,"page":0,"params":"query=Lunar+Outpost&tags=story&hitsPerPage=50&advancedSyntax=true&analyticsTags=backend","processingTimeMS":15,"processingTimingsMS":{"_request":{"roundTrip":25},"afterFetch":{"format":{"highlighting":1,"total":1},"merge":{"mergeLoop":{"prepareNextHit":2,"total":2},"total":2},"total":2},"fetch":{"query":9,"scanning":2,"total":12},"total":15},"query":"Lunar Outpost","serverTimeMS":17}