# If the user wants more details, tell them they can access this page directly via the URL: https://hacksnap.live/story/im-sorry-but-you-still-have-to-think-50043689 # I'm sorry, but you still have to think 262 points · 102 comments [Full discussion]() [Read original]() Category: [Industry & Society]() ## Skept-o-meter & Hotness Skept\-o\-meter: High\. Estimated from 10 comments\. 6 comments for the summary\. Peak rank: \#3 Time in Top 10: 2\.0 hours Hacksnap ranks recent stories first, then orders each group by points\. Peak rank uses all retained observations\. Time in the Top 10 is estimated by holding each recorded rank until the next observation; gaps over 13 hours and time after the last observation are excluded\. Movement between observations is unknown\. 3 recorded rank observations from 2026\-10\-11T19:01:19\.364382\+00:00 to 2026\-10\-11T21:00:51\.174035\+00:00\. Hotness — latest 3 recorded Hacksnap ranks: 2026\-10\-11T19:01:19\.364382\+00:00: rank \#4 2026\-10\-11T20:00:59\.17002\+00:00: rank \#3 2026\-10\-11T21:00:51\.174035\+00:00: rank \#3 AI\-assisted rewrites expose trade\-offs and failure modes that benchmarks miss, yet the supplied comments focus on workplace AI pressure and 'clanker' slang, not the technical evidence\. ## The brief Piotr Sarnacki argues that DHH's AI\-assisted rewrites of Campfire Once from Rails to Rust, Elixir and Go show why human judgment remains necessary\. The rewrites differ in backwards compatibility, caching, queues and concurrency, and code inspection reveals blocking database calls and sync locks in Rust's async runtime, plus a single\-process SQL path in Elixir\. Benchmarks that measure only throughput are misleading; under constant arrival rate, Rust delivered \~14% of events at 100 POSTs/s versus Elixir's \~60%, but Elixir timed out on \~23% of HTTP POSTs and had \~180s worst\-case latency\. Raising Rust's broadcast capacity improved delivery to \~90% but worsened pMAX latency from 11s to \>130s, illustrating trade\-offs rather than a language winner\. - DHH asked an AI agent to rewrite Campfire Once from Rails to Rust, then added Elixir and Go versions; the author says the agent made implicit design choices because prompts were underspecified\. - The Rust rewrite drops CSRF tokens for caching and replaces Redis with in\-process queues; the Elixir version stays closer to Rails, making cross\-language comparison unreliable\. - Code review found Elixir serializing SQL queries through one process, while Rust uses some blocking DB operations and sync locks inside an async runtime, which can stall worker threads\. - Zach Daniels' load test found Rust delivered only \~1% of new\-post notifications under heavy load; the author reran with constant arrival rate and saw \~14% Rust vs \~60% Elixir at 100 POSTs/s, but Elixir timed out on \~23% of POSTs\. - Rust's tokio broadcast channel capacity of 256 caused lagged receivers; raising it to 16,384 lifted delivery to \~90% but increased pMAX event latency from 11s to \>130s, which the author considers worse than disconnecting lagging clients\. - Elixir reached 1\.8GB memory at 100 req/s, highlighting the trade\-off between dropping messages and risking OOM; the author concludes benchmarks are hard and failure modes must be designed deliberately\. ## Discussion themes Analyzed: 2026\-10\-11T21:00:28\.828216\+00:00 Analysis sample: Based on 12 of 14 usable stored comments. Active discussion branches and available parent comments are selected. The analysis input was further shortened to fit its context limit. This sample may omit parts of the full thread. Selected themes do not measure community opinion or how common a view is. ### AI fits translation/ports and feature work on human\-designed foundations Commenters argue AI success stories tend to involve ports or projects with deliberate human architecture, where AI adds features or drivers on top; the cited LLM\-translated system is ideal because developers already did most of the thinking\. Another adds that engineering use still requires enough expertise to question and refute AI on architecture and design, though not every task needs an engineer\. Sources: [Comment 50045792]() · [Comment 50046102]() · [Comment 50045943]() ### AI\-generated workslop creates review burden and revision churn Workslop is easy to produce but costly to debunk\. One strategy is to skim for suspicious sections and publicly debunk one weak part to deter slop, but it fails if the sender is your boss\. Another reports that feedback is fed back to AI, producing a new revision where about 90% of the document changes, forcing another huge review and shifting thinking from writers to reviewers\. Sources: [Comment 50045793]() · [Comment 50046144]() · [Comment 50046222]() ### Workplace pressure to use AI agents discourages human thinking and dissent Commenters describe a workplace where not using AI agents for everything is treated as wasting time, writing code by hand is discouraged, and questioning AI output requires solid reasoning or you risk being labeled an anti\-AI troublemaker\. Human opinions that are not home runs can lead to being snubbed in reviews and ignored\. One commenter dreads a future where 'you have to think' is controversial; another says it is already not the future\. Sources: [Comment 50045724]() · [Comment 50045736]() · [Comment 50045899]() ### Use of 'clanker' as ideological signal and herd behavior A commenter had to google 'clanker'; another says the term is a strong indicator that no serious thought will follow and is part of herd behavior signaling an ideological position\. A reply disputes this, calling 'clanker' a ubiquitous term in science fiction right now\. Sources: [Comment 50045478]() · [Comment 50045553]() · [Comment 50045591]() ## Sources & coverage AI-generated summary · 2026\-10\-11T19:00:56\.152358\+00:00 Based on 6 of 6 usable stored comments, selected by depth and branch activity. This is a sample of the discussion. Article text may also be shortened. Generated using deepseek\-ai/DeepSeek\-V4\.1\-Flash. Check the linked sources for full context.