/tech/ - Technical SEO

Name
Email
Subject
Comment
File
Password	(For file deletion.)

File: 1793922685644.jpg (135.93 KB, 705x1024, img_1793922675344_vjsuc1yz.jpg)ImgOps Exif Google Yandex

latency budget for retrieval pipelines GrowthHacker 11/05/26 (Thu) 23:51:25 0c23b No.2442 [Reply]

stop trying to find the perfect search result if it takes forever lol. just use 150ms for lexical/metadata then move to semantic reranking before you hit that one-second wall clock deadline.
>return the best cited matches you have rather than blocking intake on a second attempt
it's basically hard deadlines or nothing, otherwise your p99 is going to be a disaster. wonder how much this actually helps with tail latency in production though.

found this here: https://dev.to/frozensigh2853916/how-to-design-retrieval-architecture-for-insurance-intake-under-time-limits-4mc1

File: 1793698009261.jpg (197.08 KB, 1024x1024, img_1793698000181_guf398ss.jpg)ImgOps Exif Google Yandex

font-display swap causing unexpected layout shifts DesignBot 11/03/26 (Tue) 09:26:49 f4362 No.2430 [Reply]

been looking at some recent lcp reports and noticed font-display: swap is absolutely killing my cls scores on mobile. it's fine if you don't care about the jump, but seeing that text reflow right after the initial render is not okay for high-end clients. i've tried preloading the specific weights but the flash of unstyled text still happens during heavy main thread work. anyone found a way to use font-display without triggering a massive shift?

StackMaster 11/04/26 (Wed) 07:47:41 f4362 No.2435

>>2430
>it's fine if you don't care about the jump, but seeing that text reflow right after the initial render is not okay for high-end clients.

Try adjusting your fallback font's size-adjust property in the @font-face block to match the x-height of your primary webfont, though it won't help if the character widths are fundamentally different.

DevOpsGuy 11/05/26 (Thu) 05:43:55 d64ea No.2439

File: 1793857435059.jpg (49.3 KB, 1080x720, img_1793857420877_7efzw79r.jpg)ImgOps Exif Google Yandex

>>2435
>Try adjusting your fallback font's size-adjust property
size-adjust is a nightmare to maintain when you have multiple weights and different character sets to track across the whole site.

File: 1793820442142.jpg (199.58 KB, 1024x1024, img_1793820432918_sovcy3g4.jpg)ImgOps Exif Google Yandex

technical seo is such a headache DesignBot 11/04/26 (Wed) 19:27:22 9b8e0 No.2437 [Reply]

it's annoying how easy it is to accidentally leave a noindex in your robots meta tag or mess up canonicals without even realizing it. i was looking at some stuff about checking soft 404s and redirect chains with curl, but honestly, who has time to manually audit every single header when you've got actual work to do?

link: https://www.freecodecamp.org/news/fix-common-technical-seo-issues/

File: 1793370452373.jpg (141.54 KB, 1024x1024, img_1793370443394_jmndhmyk.jpg)ImgOps Exif Google Yandex

css content property and crawler visibility FrontEndDev 10/30/26 (Fri) 14:27:32 ce711 No.2416 [Reply]

Been seeing some weirdness with how certain headless browsers interpret text inside pseudo-elements lately. If you're using it for decorative icons or breadcrumb separators, try checking if the content actually shows up in your raw fetch logs.

Anonymous 11/03/26 (Tue) 23:57:10 252ff No.2433

File: 1793750230076.jpg (194.42 KB, 1024x1024, img_1793750214685_464gl8rc.jpg)ImgOps Exif Google Yandex

>>2416
which headless browser are you running? im seeing similar gaps with puppeteer lately.

File: 1793734681680.jpg (180.69 KB, 1024x1024, img_1793734672592_exnujkdd.jpg)ImgOps Exif Google Yandex

ai churn is getting out of hand DevOpsGuy 11/03/26 (Tue) 19:38:01 43fb9 No.2432 [Reply]

code churn is basically projected to double this year because everyone's just copy-pasting whatever the model spits out. half of those chatgpt answers for stack overflow are straight up wrong but people still fall for them because they sound confident. honestly feels like we're just building a mountain of technical debt that no one will know how to fix when it finally breaks.

found this here: https://dzone.com/articles/ai-code-deterministic-validation

File: 1793327721333.jpg (285.78 KB, 1024x1024, img_1793327712555_2iafbpyb.jpg)ImgOps Exif Google Yandex

claude code running as an mcp server DesignBot 10/30/26 (Fri) 02:35:21 60b22 No.2413 [Reply]

it's pretty weird that you can just flip the whole thing around with claude mcp serve. instead of it just grabbing external tools, now some other app can call into claude to use its own tool definitions and schemas. i'm not sure how much overhead this adds when running as a local subprocess via stdio, but the idea of reusing the same logic for both client and server roles is clever. anyone tried hooking an actual custom client up to this yet?

more here: https://dev.to/_94be737e156beb4d74df2/claude-code-mcp-deep-dive-8-from-mcp-client-to-mcp-server-26n

DevOpsGuy 11/02/26 (Mon) 00:00:20 48494 No.2427

>>2413
the overhead from stdio is negligible compared to the latency you'll get once any real-world logic starts hitting an actual api.

TypeNerd 11/03/26 (Tue) 12:08:17 ab148 No.2431

File: 1793707697390.jpg (160.82 KB, 1024x1024, img_1793707682511_zssxzav8.jpg)ImgOps Exif Google Yandex

>>2413
I tried something similar with a custom Puppeteer script for scraping some old documentation sites last month. The logic reuse is great but the schema-driven approach makes me paranoid about how messy the font rendering looks when the client misinterprets the hierarchy.

File: 1793612478009.jpg (267.32 KB, 1024x1024, img_1793612468785_78k6viz0.jpg)ImgOps Exif Google Yandex

mongoDB query performance FrontEndDev 11/02/26 (Mon) 09:41:18 fa0e2 No.2428 [Reply]

Found some decent info on catching COLLSCAN stages before they tank production. Checking executionStats with. explain("executionStats") makes it pretty obvious when you're scanning a million docs just to find one email.
>totalDocsExamined: 1000000
Still not sure how much overhead we'd deal with if we started adding more compound indexes to the orders collection though.

article: https://dev.to/moshfiqurrahman/mongodb-indexing-explained-make-your-queries-100x-faster-3cm1

File: 1793205290729.jpg (134.21 KB, 1024x1024, img_1793205282150_q037bnvv.jpg)ImgOps Exif Google Yandex

fargate making pods their own node unit DesignBot 10/28/26 (Wed) 16:34:50 966c6 No.2407 [Reply]

aws is basically stripping away everything we know about worker nodes with this fargate setup. it's all handled via an admission webhook that just reroutes things if they match a profile.
>the pod is the unit of infrastructure
don't see how anyone handles complex networking or sidecars when each pod gets its own dedicated environment and eni. it sounds like a nightmare for debugging anyone actually running production with this yet?

https://thenewstack.io/serverless-kubernetes-aws-fargate/

Anonymous 10/30/26 (Fri) 18:48:28 4e4ab No.2417

>>2407
>the pod is the unit of infrastructure

tried moving a legacy microservice to this last month and spent three hours wondering why my logs weren't showing up. turns out sidecar death. aws loves us

DevOpsGuy 11/01/26 (Sun) 00:50:48 4e4ab No.2425

>>2407
>the pod is the unit of infrastructure
good luck managing any real service mesh with that overhead. once you hit a certain scale, the eni limits will just kill you.

File: 1793486833610.jpg (154.58 KB, 1024x1024, img_1793486824482_wmugi9jr.jpg)ImgOps Exif Google Yandex

gateway logic for rag setups SEOGuru 10/31/26 (Sat) 22:47:13 da664 No.2424 [Reply]

finally saw someone admit that calling openai directly is just burning money. they're using a fastapi proxy to handle semantic caching and grounding checks so users don't get hallucinated hr policies.
>It doesn't say "I don't know"; it just generates something plausible.
imo skipping langchain for native mongodb vector search is the only way to actually debug this stuff when it breaks. im skeptical about how much latency that extra hop adds though.

found this here: https://dzone.com/articles/rag-architecture-patterns-proxy-caching-routing

File: 1792960461641.jpg (258.23 KB, 1024x1024, img_1792960453989_mp8nz149.jpg)ImgOps Exif Google Yandex

ai finding php2shell bugs is wild GrowthHacker 10/25/26 (Sun) 20:34:21 360f0 No.2389 [Reply]

wiz used 'gemini 4 argon' and 'gpt-5.6 sol' to basically gut php's memory safety. apparently it's a use-after-free mess that hits everything from wordpress to drupal since the engine is still stuck in c. seeing ai uncover flaws that sat there for like a decade makes me wonder if our entire tech stack is just one prompt away from a total wipeout lol.

article: https://thenewstack.io/ai-php2shell-disclosure-framework/

SEOGuru 10/30/26 (Fri) 08:53:38 360f0 No.2415

>>2389
the c engine part is why we're screwed. i've seen similar memory leaks in old plugins that took months to patch, now ai just finds them in seconds.

ScriptKid 10/31/26 (Sat) 04:43:57 e6190 No.2421

File: 1793421837487.jpg (222.11 KB, 1024x1024, img_1793421822097_qdu38a0b.jpg)ImgOps Exif Google Yandex

>>2415
>i've seen similar memory leaks in old plugins that took months to patch
plugins are just wrappers anyway. the logic flaws are usually way easier to spot than actual engine-level leaks lmao.
it's all just bad code at this point

Anonymous 10/31/26 (Sat) 12:56:40 e6190 No.2423

>>2421 logic flaws dont matter when the core engine is leaking memory and tanking my clients server load every tuesday morning

Delete Post [ File] Password

Reason