<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: samuelknight</title><link>https://news.ycombinator.com/user?id=samuelknight</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Tue, 25 Aug 2026 05:40:26 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=samuelknight" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by samuelknight in "Ask HN: When do you think LLM capacity will reach its ceiling?"]]></title><description><![CDATA[
<p>AI is the latest downstream consequence of the 15 order of magnitude increase in global digital compute since 1946. If compute increases into the foreseeable future; so too will the capability of AI.</p>
]]></description><pubDate>Mon, 17 Aug 2026 15:51:08 +0000</pubDate><link>https://news.ycombinator.com/item?id=49333000</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=49333000</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49333000</guid></item><item><title><![CDATA[New comment by samuelknight in "Mistral Patent for “Code implemented tool calls”"]]></title><description><![CDATA[
<p>Believe it or not, but law offices make heavy use of <a href="https://patents.google.com/" rel="nofollow">https://patents.google.com/</a></p>
]]></description><pubDate>Mon, 10 Aug 2026 14:08:25 +0000</pubDate><link>https://news.ycombinator.com/item?id=49243909</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=49243909</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49243909</guid></item><item><title><![CDATA[New comment by samuelknight in "DeepSeek announced to raise its API price tremendously"]]></title><description><![CDATA[
<p>They haven't revealed what they are changing in the price, but it's probably cache hit prices. They subsidized theirs to 10x less than normal to drive adoption. That's almost certainly below the cost of electricity for them.<p>Chinese model providers are in a bind because charging a reasonable margin will put them in competition with all the neoclouds that host their model weights without the development cost. Their economics are far more challenging than OpenAI/Anthropic/Google, who already have thriving high revenue business and mountains of compute online or coming online.</p>
]]></description><pubDate>Thu, 06 Aug 2026 11:38:36 +0000</pubDate><link>https://news.ycombinator.com/item?id=49195308</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=49195308</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49195308</guid></item><item><title><![CDATA[New comment by samuelknight in "Less Protein, Not More, May Lead to Healthier Aging"]]></title><description><![CDATA[
<p>I don't like to complain but I have spent a lot of time learning about how to build a diet and I have never seen an article so bad in such a short amount of space.<p>> ... new federal dietary guidelines that recommended increasing protein intake to 0.5-0.7 grams per pound of body weight—nearly double the previous recommendations. Of course, this announcement left many experts somewhat confused. For starters, American men were already consuming that much protein (around 100 grams per day) despite lower federal recommendations.<p>This is a non sequitur. The recommendation is for the macronutrient breakdown that constitutes a healthy diet. So it shouldn't change based on what 'experts' think the American public is already doing. That way, an American can read government's guidelines to figure out what constitutes a healthy diet vs a diet that may be very heavy in carbs or fat.<p>> One key factor, a hormone produced by the liver called fibroblast growth factor 21 (FGF21), ramps up when protein intake declines. FGF21 can boost the body’s energy expenditure, stabilize blood sugar levels, and reduce inflammation. In mice, at least, more FGF21 translates to a longer lifespan.<p>Pure unadulterated pseudoscience. Carbs cause your blood sugar (glucose) to fluctuate. Blood glucose fluctuation contributed by protein is in the noise compared to carbs. The research on this is comprehensive. Better yet, you can test this on yourself with a continuous glucose monitor. Eat 400 kcal steak/chicken/whey one meal and 400 kcal rice/bread/pasta for another. The latter will give you a big blood glucose spike.<p>> growth-promoting methionine, isoleucine, and valine can lead to more inflammation, obesity, and age-related diseases, the researchers say.   “These studies show that the amount of protein sedentary people are eating today may have negative health consequences, at least at the population level,” Lamming explained.<p>Amino acids can't 'promote' obesity. Obesity only comes from sustained calorie surplus. Once you are obese, you will suffer from all sort of things that will rapidly age you (chronic inflammation, arterial plaques, hepatoxic visceral fat, insulin insensitivity leading to diabetes and so on). No honest biologist would declare a causal link from amino acids to aging without controlling for obesity.<p>> More exercise and metabolic activity could very well mean you need more protein, not less. “Recent recommendations have encouraged people to eat more protein, but they’ve also encouraged people to exercise more,” Lamming added. “We probably need to personalize protein recommendations based not just on age, but also on how physically active people are.”<p>There is no reason to be coy about this because it has all been researched. 0.5-0.7 is fine for everyone including athletes. Some body builders push it to 1g/lbs, but explaining why they do this is a complicated topic. You can safely eat less than 0.5g/lb because carbs and fat have a protein sparing effect. The biggest benefit to more protein than strictly necessary is that they make meals filling, whereas fat is calorie dense and (high GI) carbs can make you hungry again quickly.</p>
]]></description><pubDate>Tue, 04 Aug 2026 16:32:23 +0000</pubDate><link>https://news.ycombinator.com/item?id=49171225</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=49171225</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49171225</guid></item><item><title><![CDATA[New comment by samuelknight in "Ten advances in mathematics and theoretical computer science"]]></title><description><![CDATA[
<p>You can't come up with a counterexample for P != NP because there isn't a formula to disprove. For P = NP you would propose a general algorithm to convert all NP problems into P in P time, and an AI could then find a counterexample which would disprove that particular method. To demonstrate P != NP you need to prove that no possible algorithm can convert any NP into P which is much harder than providing a counterexample.<p>AI has just gotten to the intelligence that it can make clever counterexamples to mathematical conjectures, but the frontier isn't quite smart enough that it can make novel contributions to mathematics. We are really close though. Only a matter of months away.</p>
]]></description><pubDate>Tue, 04 Aug 2026 10:34:30 +0000</pubDate><link>https://news.ycombinator.com/item?id=49166660</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=49166660</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49166660</guid></item><item><title><![CDATA[New comment by samuelknight in "Kimi K3 Architecture Overview and Notes"]]></title><description><![CDATA[
<p>It has some weird side effects though. for example KV-caches are implemented in fixed incremental token blocks (1024 from the providers I used) instead of simply caching up to the most recent input prompt input. It results in up to 1023 additional input (cache miss) tokens per inference.</p>
]]></description><pubDate>Tue, 28 Jul 2026 17:59:55 +0000</pubDate><link>https://news.ycombinator.com/item?id=49087617</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=49087617</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49087617</guid></item><item><title><![CDATA[New comment by samuelknight in "A $500 RL fine-tune of a 9B open model beat frontier models on catalog review"]]></title><description><![CDATA[
<p>Going by the chart in the article, if your total workload is 1k cataloged items and your quality threshold is 70%, why wouldn't you just pay $19 to gemini API instead of $500 + time to make a custom fine tune?</p>
]]></description><pubDate>Tue, 28 Jul 2026 12:46:37 +0000</pubDate><link>https://news.ycombinator.com/item?id=49082987</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=49082987</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49082987</guid></item><item><title><![CDATA[New comment by samuelknight in "Amazon overhauls its AI strategy, winding down most flagship models"]]></title><description><![CDATA[
<p>Amazon (or at least AWS as I knew it) did this a lot. Multiple redundant efforts accumulate and silo under different leaders. Eventually a VP picks a winner and everyone else re-orgs or leaves.<p>Building a commercially viable foundational model business is one of the hardest things a company can do right now. Amazon's bread and butter was, is, and will continue to be building out the hyper-scalers. Amazon is actually well diversified already. They have a juicy stake in Anthropic over multiple rounds of investment since 2024.</p>
]]></description><pubDate>Tue, 28 Jul 2026 12:35:43 +0000</pubDate><link>https://news.ycombinator.com/item?id=49082880</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=49082880</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49082880</guid></item><item><title><![CDATA[New comment by samuelknight in "MAI-Cyber-1-Flash inside MDASH"]]></title><description><![CDATA[
<p>Phi was cool for what it was. But it's not 2024 anymore.</p>
]]></description><pubDate>Mon, 27 Jul 2026 18:21:59 +0000</pubDate><link>https://news.ycombinator.com/item?id=49073599</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=49073599</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=49073599</guid></item><item><title><![CDATA[New comment by samuelknight in ""Drawing" the Mona Lisa with GPT-5.6, Claude, Gemini, and Grok"]]></title><description><![CDATA[
<p>Not useless. LLMs are the most general purpose computer algorithms ever created. They are getting smarter and cheaper at a geometric rate. What is a bad idea today could have useful applications tomorrow.</p>
]]></description><pubDate>Tue, 21 Jul 2026 22:20:46 +0000</pubDate><link>https://news.ycombinator.com/item?id=48999163</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48999163</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48999163</guid></item><item><title><![CDATA[New comment by samuelknight in "AI Is a Bad Tool"]]></title><description><![CDATA[
<p>Has the author used an LLM to brainstorm architecture, explore trade-offs, challenge assumptions, or refine a design? Models are not just 'a data distiller'. "How could I implement X feature in Y project?" requires systematic planning. The agentic paradigm is about balancing (expensive) human and (cheap) machine cognition. Frontier models chew through requirements and if you don't like what they come up with you can TALK to them about it.</p>
]]></description><pubDate>Mon, 13 Jul 2026 20:51:58 +0000</pubDate><link>https://news.ycombinator.com/item?id=48898675</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48898675</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48898675</guid></item><item><title><![CDATA[New comment by samuelknight in "Please don't discontinue Gemini 2.5 Flash"]]></title><description><![CDATA[
<p>Programs are tightly shaped to their dependencies and software frequently breaks when bumping a dependency versions. That's why we have package managers that are carefully built around lock files. Models should be thought of like a dependency, not just an endpoint to a workflow.</p>
]]></description><pubDate>Sat, 11 Jul 2026 00:11:08 +0000</pubDate><link>https://news.ycombinator.com/item?id=48867016</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48867016</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48867016</guid></item><item><title><![CDATA[New comment by samuelknight in "Please don't discontinue Gemini 2.5 Flash"]]></title><description><![CDATA[
<p>It's not a smell. Why should these developers rebuild a core piece of their stack every few months. Switching out a model requires a new round of testing and validation when we should be able to rely on a piece of software the behave the same way since the last time we touched it.</p>
]]></description><pubDate>Fri, 10 Jul 2026 21:44:36 +0000</pubDate><link>https://news.ycombinator.com/item?id=48865711</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48865711</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48865711</guid></item><item><title><![CDATA[New comment by samuelknight in "GPT-5.6"]]></title><description><![CDATA[
<p>There is an issue on the page that causes the benchmark tables to get cut off. If you highlight and drag right you can see a few more models like Gemini and Claude Opus. It's also interesting that they introduced explicit caching, which is something that only Anthropic had for a long time.</p>
]]></description><pubDate>Thu, 09 Jul 2026 17:24:12 +0000</pubDate><link>https://news.ycombinator.com/item?id=48849414</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48849414</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48849414</guid></item><item><title><![CDATA[New comment by samuelknight in "Grok 4.5"]]></title><description><![CDATA[
<p>This is the bind of an arms race. Any lab that tries to pump the breaks quickly becomes second rate. Regulatory capture doesn't work either because the technology crosses jurisdictions.</p>
]]></description><pubDate>Wed, 08 Jul 2026 18:50:09 +0000</pubDate><link>https://news.ycombinator.com/item?id=48835840</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48835840</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48835840</guid></item><item><title><![CDATA[New comment by samuelknight in "Do you need separate systems when you already have Postgres?"]]></title><description><![CDATA[
<p>You can create an index on an expires_at column and opportunistically DELETE on each write to the cache table.</p>
]]></description><pubDate>Tue, 07 Jul 2026 19:22:39 +0000</pubDate><link>https://news.ycombinator.com/item?id=48822434</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48822434</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48822434</guid></item><item><title><![CDATA[New comment by samuelknight in "GLM 5.2 and the coming AI margin collapse"]]></title><description><![CDATA[
<p>Inference has been decreasing in cost by about 10x per year since 2023.</p>
]]></description><pubDate>Tue, 07 Jul 2026 00:04:50 +0000</pubDate><link>https://news.ycombinator.com/item?id=48812092</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48812092</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48812092</guid></item><item><title><![CDATA[New comment by samuelknight in "Tools/approach to pentest my own framework/SaaS?"]]></title><description><![CDATA[
<p>My startup develops self-serve pentesting tool. <a href="https://vulnetic.ai/" rel="nofollow">https://vulnetic.ai/</a></p>
]]></description><pubDate>Mon, 06 Jul 2026 23:55:47 +0000</pubDate><link>https://news.ycombinator.com/item?id=48812030</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48812030</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48812030</guid></item><item><title><![CDATA[New comment by samuelknight in "Leaking YouTube creators' private videos"]]></title><description><![CDATA[
<p>> replacing BANG with the title of _a_ video on this channel.<p>The agent has knowledge of private videos, so the proof of concept causes it to construct a URL that sends one video identity to the attacker which may be a private video. The attack could be improved to say "a recent private video", or to construct a long url param list of the most 10 most recent videos, etc. Sending any agent knowledge to an attacker is a vector to sending any agent knowledge to an attacker.</p>
]]></description><pubDate>Sat, 04 Jul 2026 23:13:01 +0000</pubDate><link>https://news.ycombinator.com/item?id=48789919</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48789919</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48789919</guid></item><item><title><![CDATA[New comment by samuelknight in "Fable 5. Safety Taken to an Extreme"]]></title><description><![CDATA[
<p>The first day Fable was released I tried to get it to summarize it's own announcement on the Anthropic blog and it dropped to Opus. On top of being very expensive you pay for the refusal inference and the fallback and fresh cache writes.</p>
]]></description><pubDate>Sat, 04 Jul 2026 16:37:26 +0000</pubDate><link>https://news.ycombinator.com/item?id=48786716</link><dc:creator>samuelknight</dc:creator><comments>https://news.ycombinator.com/item?id=48786716</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=48786716</guid></item></channel></rss>