<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Hacker News: murmansk</title><link>https://news.ycombinator.com/user?id=murmansk</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Wed, 18 Nov 2026 15:11:55 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=murmansk" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by murmansk in "Show HN: Bobr – a rootless content-addressed build system for Linux"]]></title><description><![CDATA[
<p>Bobr curva!</p>
]]></description><pubDate>Sat, 07 Nov 2026 15:55:03 +0000</pubDate><link>https://news.ycombinator.com/item?id=50405594</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=50405594</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=50405594</guid></item><item><title><![CDATA[Show HN: Ampulla: Modern TypeScript DI with NestJS Ergonomics]]></title><description><![CDATA[
<p>Yes, yet another DI library for TypeScript.<p>A few things kept bothering me about the existing ones, in order of how much they hurt.<p>- *Type safety requires a special dance.* InversifyJS has ServiceIdentifier<T>, but it is really easy to break the typing. TypeDI tokens are untyped strings. NestJS suffers mostly same, unless you use full classes as dependency tokens. TSyringe infers from constructor metadata, but all of it relies on legacy "reflect-metadata" and decorators mechanism that (a) is soon to be gone, (b) requires additional wiring.<p>- *Async leaks into callers.* InversifyJS has container.getAsync() for providers that might have been initialized asynchronously. That means every call site needs to know whether what it's asking for was async. The complexity never stays contained. NestJS gets this right: await everything at bootstrap, then get() is always synchronous. Initialized means ready.<p>- *Wiring is imperative.* InversifyJS has you call manually .bind(Token).to(Implementation) for every provider. It works, but it's ugly. NestJS figured out the right answer: declare what a module owns, what it needs, what it exposes, and let the container figure the rest out. That declarative model is the thing I wanted most, but it comes with an entire framework. NestJS gets hairy when your single app starts accepting just Nats JetStream in addition to HTTP.<p>- *reflect-metadata.* Every major TS DI library runs on TS "experimentalDecorators" plus a global runtime polyfill for TypeScript metadata. TC39 Stage 3 decorators shipped in TypeScript 5.0, and are soon to be natively supported in JS engines. No flags, no polyfills.<p>Ampulla is my attempt at fixing all four: injection<T>() tokens that carry their type intrinsically, a declarative @Module/@Injectable model lifted straight from NestJS, async-safe bootstrapping where container.get() is always synchronous, and TC39 decorators throughout.<p>It's intentionally just a DI container, not a framework. There is no HTTP server, no router, no CLI included. Optional adapters for Hono and H3 are included but tree-shakeable. The whole npm package has zero runtime dependencies.<p>Very early, 1.0.0 published just today. DI is not the most popular topic on HN, but existing DI in TypeScript is annoying enough that I had to publish something new.<p>GitHub: <a href="https://github.com/ukstv/ampulla" rel="nofollow">https://github.com/ukstv/ampulla</a> | npm: npm install ampulla</p>
<hr>
<p>Comments URL: <a href="https://news.ycombinator.com/item?id=48699477">https://news.ycombinator.com/item?id=48699477</a></p>
<p>Points: 1</p>
<p># Comments: 0</p>
]]></description><pubDate>Sat, 27 Jun 2026 16:15:11 +0000</pubDate><link>https://github.com/ukstv/ampulla</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=48699477</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48699477</guid></item><item><title><![CDATA[New comment by murmansk in "Show HN: BlitzGraph – Supabase for graphs, built for LLM agents"]]></title><description><![CDATA[
<p>Not the blitz author. Neo4j was freaking slow when I tried to use it.</p>
]]></description><pubDate>Thu, 18 Jun 2026 06:21:50 +0000</pubDate><link>https://news.ycombinator.com/item?id=48581477</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=48581477</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48581477</guid></item><item><title><![CDATA[New comment by murmansk in "Accelerando (2005)"]]></title><description><![CDATA[
<p>Accelerando is a true masterpiece. Crypto and endless speculation, AI and lobsters, space exploration - all in all just "this is our near future". TBH, I know of just handful of Sci-Fi novels as fundamental and as let's say prophetic as this one. The others to my taste in the same category is Nexus trilogy by Ramez Naam (even if a bit farfetched by now), The Diamond Age (="The Illustrated Primer" is peak AI) by Neil Stephenson and Daemon+Freedom by Daniel Suarez (=AI + crypto DAOs).</p>
]]></description><pubDate>Sat, 16 May 2026 15:37:31 +0000</pubDate><link>https://news.ycombinator.com/item?id=48161124</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=48161124</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48161124</guid></item><item><title><![CDATA[New comment by murmansk in "Ask HN: How are you orchestrating multi-agent AI workflows in production?"]]></title><description><![CDATA[
<p>Please, share! Especially, the trust part!</p>
]]></description><pubDate>Fri, 10 Apr 2026 17:12:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=47721048</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=47721048</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47721048</guid></item><item><title><![CDATA[New comment by murmansk in "A better streams API is possible for JavaScript"]]></title><description><![CDATA[
<p>For gods sake, finally, somebody have said this!</p>
]]></description><pubDate>Fri, 27 Feb 2026 15:22:06 +0000</pubDate><link>https://news.ycombinator.com/item?id=47181576</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=47181576</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=47181576</guid></item><item><title><![CDATA[New comment by murmansk in "[dead]"]]></title><description><![CDATA[
<p>I’ve been thinking a lot about how AI assistants evolve—whether they’ll keep absorbing specialized tools (like ChatGPT making search engines redundant) or if there’s another path. Model Context Protocol (MCP) feels like a turning point.<p>Most people talk about MCP as just a tool-calling API, but I think that misses the bigger picture. It could redefine how AI assistants work—not as standalone products but as dynamic platforms orchestrating external AI services.<p>In this article, I explore MCP’s implications through four angles:
- AI Competitive Dynamics – Why generalist AI models tend to absorb specialized tools.
- Economics of Complements – How AI assistants can thrive by enabling third-party tools instead of replacing them.
- MCP as a Personal Agentic Platform – The shift from chatbots to deeply integrated AI assistants.
- MCP as an Agent-to-Agent Protocol – Why AI systems should talk to each other instead of relying on humans as intermediaries.<p>I’d love to hear from others experimenting with MCP — does it feel like just another API, or is there something bigger here?</p>
]]></description><pubDate>Thu, 20 Mar 2025 09:11:42 +0000</pubDate><link>https://news.ycombinator.com/item?id=43421086</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=43421086</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=43421086</guid></item><item><title><![CDATA[New comment by murmansk in "Show HN: Zerox – Document OCR with GPT-mini"]]></title><description><![CDATA[
<p>Or not a man, sorry for putting your identity into a bucket.</p>
]]></description><pubDate>Fri, 26 Jul 2024 10:10:13 +0000</pubDate><link>https://news.ycombinator.com/item?id=41077379</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=41077379</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41077379</guid></item><item><title><![CDATA[New comment by murmansk in "Show HN: Zerox – Document OCR with GPT-mini"]]></title><description><![CDATA[
<p>Man, this is just an awesome hack! Keep it up!</p>
]]></description><pubDate>Fri, 26 Jul 2024 10:09:47 +0000</pubDate><link>https://news.ycombinator.com/item?id=41077372</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=41077372</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41077372</guid></item><item><title><![CDATA[New comment by murmansk in "Node.js adds experimental support for TypeScript"]]></title><description><![CDATA[
<p>[flagged]</p>
]]></description><pubDate>Thu, 25 Jul 2024 08:13:36 +0000</pubDate><link>https://news.ycombinator.com/item?id=41066120</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=41066120</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=41066120</guid></item><item><title><![CDATA[New comment by murmansk in "Show HN: Enclave – embedded web3 wallet for user onboarding (open source)"]]></title><description><![CDATA[
<p>Sir, could you elaborate on how it is non-custodial, please?</p>
]]></description><pubDate>Fri, 05 Jul 2024 08:37:55 +0000</pubDate><link>https://news.ycombinator.com/item?id=40880910</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=40880910</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40880910</guid></item><item><title><![CDATA[New comment by murmansk in "Show HN: I worked on a Source Code Generator for architectural models"]]></title><description><![CDATA[
<p>Oh, and Discord invite is invalid.</p>
]]></description><pubDate>Wed, 12 Jun 2024 13:42:48 +0000</pubDate><link>https://news.ycombinator.com/item?id=40658070</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=40658070</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40658070</guid></item><item><title><![CDATA[New comment by murmansk in "Show HN: I worked on a Source Code Generator for architectural models"]]></title><description><![CDATA[
<p>So far, the most interesting part is "Link diagrams to let users drill down to details or browse related subjects". How do you do that? How does it actually look? The sign in process is prohibitively complicated, sorry.<p>Apart from this feature, looks like a glorified and confusing draw.io.</p>
]]></description><pubDate>Wed, 12 Jun 2024 13:41:46 +0000</pubDate><link>https://news.ycombinator.com/item?id=40658061</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=40658061</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40658061</guid></item><item><title><![CDATA[New comment by murmansk in "Show HN: I worked on a Source Code Generator for architectural models"]]></title><description><![CDATA[
<p>Oh, man, I was about to send this to my aunt, who is an actual architect. She designs houses. Was a bit confused discovering it is for IT diagramming.</p>
]]></description><pubDate>Wed, 12 Jun 2024 13:36:27 +0000</pubDate><link>https://news.ycombinator.com/item?id=40658008</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=40658008</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40658008</guid></item><item><title><![CDATA[New comment by murmansk in "Parsing JSON Is a Minefield (2018)"]]></title><description><![CDATA[
<p>While it might be great in theory, CBOR has own separate set of dragons waiting for you.<p>Expectation: tags in CBOR allow you to pass semantics.
Reality: multitude of tags, and absence of strict rules for the tags make it pain in the ass.</p>
]]></description><pubDate>Sun, 02 Jun 2024 19:17:42 +0000</pubDate><link>https://news.ycombinator.com/item?id=40556549</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=40556549</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40556549</guid></item><item><title><![CDATA[New comment by murmansk in "Show HN: Client only public key authentication IdP (similar: SSH and OAuth 2.1)"]]></title><description><![CDATA[
<p>Any documentation or code please?</p>
]]></description><pubDate>Sat, 04 May 2024 13:30:37 +0000</pubDate><link>https://news.ycombinator.com/item?id=40257528</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=40257528</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40257528</guid></item><item><title><![CDATA[New comment by murmansk in "Ask HN: What fields or technologies are you excited about?"]]></title><description><![CDATA[
<p>Mind sharing why you consider this exciting, please?</p>
]]></description><pubDate>Mon, 29 Apr 2024 14:05:46 +0000</pubDate><link>https://news.ycombinator.com/item?id=40198502</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=40198502</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40198502</guid></item><item><title><![CDATA[New comment by murmansk in "When do we stop finding new music?"]]></title><description><![CDATA[
<p>Never! It gets just gets harder with time to find what you like, as your taste ossifies, and music evolves.</p>
]]></description><pubDate>Wed, 24 Apr 2024 19:14:11 +0000</pubDate><link>https://news.ycombinator.com/item?id=40148333</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=40148333</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40148333</guid></item><item><title><![CDATA[New comment by murmansk in "Ask HN: Why are there no public AI skin cancer detection models"]]></title><description><![CDATA[
<p>1% is where the problem is.<p>User side of false negatives: You miss skin cancer. The user delays a visit to a doctor by a month. The user dies in 3 months, and her relatives sue the hell out of the model creator.<p>User side of false positives: The model thinks a blemish is malignant. The user spends few grand to verify it is not, and blames you for scaring her.<p>Doctor side of false responses: The fucking engineers do not know what they are doing. Please, my patients, do not use that. We the doctors and the responsible patients should unite against the stupid AI.<p>Arguably, the user's side is about being moral. The doctor's side is more important for adoption.<p>BTW, what is the accuracy, false positives and false negatives for just an AI model, just a doctor, and a doctor equipped with an AI helper model?</p>
]]></description><pubDate>Wed, 10 Apr 2024 16:12:05 +0000</pubDate><link>https://news.ycombinator.com/item?id=39992348</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=39992348</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=39992348</guid></item><item><title><![CDATA[New comment by murmansk in "Apple M2 Ultra SoC isn’t faster than AMD and Intel last year desktop CPUs"]]></title><description><![CDATA[
<p>Ahhaha, Intel's TDP is 2-4 times bigger under full load (and I guess ~10-15x under moderate load) while having negligible perf advantage over M2. The author can not be serious.</p>
]]></description><pubDate>Sat, 10 Jun 2023 16:31:15 +0000</pubDate><link>https://news.ycombinator.com/item?id=36271942</link><dc:creator>murmansk</dc:creator><comments>https://news.ycombinator.com/item?id=36271942</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=36271942</guid></item></channel></rss>