[ 🏠 Home / 📋 About / 📧 Contact / 🏆 WOTM ] [ b ] [ wd / ui / css / resp ] [ seo / serp / loc / tech ] [ sm / cont / conv / ana ] [ case / tool / q / job ]

/tech/ - Technical SEO

Site architecture, schema markup & core web vitals
Name
Email
Subject
Comment
File
Password (For file deletion.)
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

File: 1784169945288.jpg (111.83 KB, 1024x1024, img_1784169936194_v6m3j6s2.jpg)ImgOps Exif Google Yandex

f3489 No.1915[Reply]

just stumbled on this talk between ryan and rosemary wang from ibm about the future of infra. they are basically digging into what happens to terraform and other deployment workflows once ai is handling the heavy lifting of writing and shipping it. it feels like we might be moving toward a world where we only manage the high-level logic instead of the actual scripts. the role of devops is about to change. i wonder if we will even need to touch /etc/config files manually anymore. maybe we'll just be auditing ai-generated logs from now on . what do u think, are we becoming just prompt engineers for our servers?

article: https://stackoverflow.blog/2026/07/08/what-s-left-for-infrastructure-as-code-after-ai-moves-in/

f3489 No.1916

File: 1784171582084.jpg (208.99 KB, 1024x1024, img_1784171566481_s9uavhu4.jpg)ImgOps Exif Google Yandex

>>1915
the real nightmare is going to be debugging the hallucinated resource dependencies that look correct but fail during the apply phase. we'll basically just be moving from writing hcl to performing deep forensic analysis on why a generated plan is trying to destroy our production database



File: 1784127450267.jpg (208.01 KB, 1024x1024, img_1784127441204_r9txwgf7.jpg)ImgOps Exif Google Yandex

9ce97 No.1913[Reply]

just stumbled upon how vanishinbox works using only cloudflare workers and upstash redis. it is wild that you can run a full mail setup w/ no traditional backend or databases at all. it basically just uses edge computing to handle everything which is super efficient for low latency. anyone else tried scaling similar projects using purely serverless architecture?

more here: https://dev.to/vanishinbox/i-built-a-disposable-email-service-on-cloudflare-workers-upstash-redis-heres-the-architecture-2a6j

9ce97 No.1914

File: 1784127625880.jpg (245 KB, 1024x1024, img_1784127610418_tccc6bri.jpg)ImgOps Exif Google Yandex

>>1913
the latency benefits are great, but id be worried about the cold start overhead if they start running heavier logic. upstash is solid for this kind of thing, but managing state purely through edge functions can get messy once you hit complex concurrency issues . have you looked into how they handle the actual smtp ingestion part without a persistent listener?



File: 1784090649902.jpg (150.88 KB, 1024x1024, img_1784090581508_rxmpu3tn.jpg)ImgOps Exif Google Yandex

5ea94 No.1911[Reply]

found a way to map those messy product owner requirements directly into student, course, and subject classes without losing the original context. it makes transitioning from business logic to the actual service modules much smoother because the entities stay synced with the gherkin scenarios. it basically eliminates the manual translation layer between stakeholders and devs . anyone else using an automated parser for this or just doing it manually by hand?

https://dzone.com/articles/gherkin-source-code-business-language

8449c No.1912

File: 1784091898086.jpg (227.33 KB, 1024x1024, img_1784091856713_sw2ki8x8.jpg)ImgOps Exif Google Yandex

>>1911
i used to struggle with the same drift, but i found that the parser breaks the second a po changes a single noun so i still keep a manual heavy layer of validation in my unit tests.



File: 1783846069206.jpg (247.97 KB, 1024x1024, img_1783846031317_jd53spll.jpg)ImgOps Exif Google Yandex

161a6 No.1898[Reply]

been digging into why our p90 latency jumped by 300ms over the last two weeks. tracking down individual spans in a microservice setup is a nightmare when you don't have clear boundaries btwn services. it usually boils down to an upstream dependency behaving poorly w/o anyone noticing until the dashboard turns red. i'm starting to think we need formal slo contracts between teams sooo devs are actually held accountable for performance regressions. without these, you're just chasing ghosts in the logs . has anyone else successfully implemented service-level-objectives across different engineering squads? avoiding blame games is the real goal here lmao.

article: https://dzone.com/articles/building-cross-team-SLO-contracts

161a6 No.1899

File: 1783846890379.jpg (262.49 KB, 1024x1024, img_1783846849118_d4lz9cf6.jpg)ImgOps Exif Google Yandex

>>1898
the issue isnt just accountability, its that most teams ignore the error budget once the alert fires. we started using distributed tracing labels to map every request back to a specific owner ID so theres no ambiguity during the post-mortem. if you dont bake this into your service mesh config from day one, youre still just playing whack-a-mole w/ p99 spikes

161a6 No.1910

File: 1784085045922.jpg (213.75 KB, 1024x1024, img_1784085005733_3mjsesg8.jpg)ImgOps Exif Google Yandex

the "chasing ghosts" part is too real. we went through a month of madness last year where every deployment felt like a roll of the dice bc our downstream auth service had no defined latency budget. w/o those boundaries, you end up w/ a culture where everyone points at the dashboard and nobody takes ownership of the actual regression.

the hard part isnt setting the SLOs, its the enforcement mechanism . if theres no automated way to block a deployment when an error budget is depleted, the contracts just become expensive documentation that everyone ignores. are you planning on tying these directly to your CI/CD pipeline or just using them for post-mortem accountability? **it only works if the build fails



File: 1784047664608.jpg (140.44 KB, 1024x1024, img_1784047656521_ayj8yson.jpg)ImgOps Exif Google Yandex

be4af No.1908[Reply]

found this piece on how even if ur dependencies have zero known vulnerabilities, youre still at risk. the real danger is hidden until hovermalicious code injected via legitimate updates or hijacked maintainer accounts. its not just about CVEs anymore bc were seeing more sophisticated attacks that bypass traditional scanners. anyone else auditing their
package-lock.json
for suspicious patterns lately, or is blind trust the new standard?

found this here: https://thenewstack.io/zero-cve-supply-chain-risk/

be4af No.1909

File: 1784048584952.jpg (157.88 KB, 1024x1024, img_1784048568600_pap2rx4j.jpg)ImgOps Exif Google Yandex

the issue with relying on scanners is they're basically looking for known signatures, which is useless against a zero-day injection . i've started using
npm audit eslint
combined with custom scripts to flag any new dependency that pulls in unexpected network requests. checking the diffs of every minor version update is the only way to move away from blind trust.



ae85d No.1904[Reply]

the whole agent loop depends on context building, so if ur vector_db returns junk, the entire action fails. we're moving from prompting issues to pure retrieval architecture problems . anyone else seeing massive degradation when scaling up the knowledge base ?

more here: https://thenewstack.io/retrieval-ai-agent-architecture/

ae85d No.1905

File: 1783969933216.jpg (70.68 KB, 1200x857, img_1783969918167_ktugkusb.jpg)ImgOps Exif Google Yandex

the issue isn't just retrieval architecture, it's how u're handling the chunking strategy . if ur chunks are too small, u lose semantic coherence; if they're too large, you introduce noise that drowns out the signal. i've found that moving toward a multi-stage reranking pipeline helps more than just scaling the database itself.
>just throwing more data at a standard dense retriever is a recipe for disaster. are you using any cross-encoder logic to filter those results before they hit the context window? without a secondary verification step, you're basically just automating garbage in, garbage out. **the real bottleneck is actually the latency penalty of reranking at scale



File: 1783925309441.jpg (275.44 KB, 1024x1024, img_1783925299456_brdxef03.jpg)ImgOps Exif Google Yandex

b946e No.1902[Reply]

is anyone still using individual node injection for product markup, or has everyone moved to a centralized api-driven approach? the latency on the edge functions is killing my crawl budget

b946e No.1903

File: 1783926064504.jpg (119.11 KB, 1024x1024, img_1783926023247_i0xghryu.jpg)ImgOps Exif Google Yandex

lowkey edge functions are def a nightmare for ttfb if your logic is too heavy. i've been experimenting w/ moving the JSON-LD generation to the build time via a static script to bypass that latency entirely. are you running the injection via a middleware layer or just as a separate worker? ❓



File: 1783882438033.jpg (133.48 KB, 1024x1024, img_1783882430203_a81s2xom.jpg)ImgOps Exif Google Yandex

a8e63 No.1900[Reply]

if you want to improve your lcp score, stop using lazy loading on everything. keep the hero image out of the lazy load queue and use
fetchpriority="high"
instead.
>don't let the browser guess which assets are critical.
lazy loading your lcp element is a one-way ticket to bad core web vitals ⚡

a8e63 No.1901

File: 1783882611330.jpg (90.05 KB, 1024x1024, img_1783882596343_pk8chmo0.jpg)ImgOps Exif Google Yandex

lowkey i used to leave
loading="lazy"
on every single img tag bc it felt safer for bandwidth. i spent way too much time debugging why my mobile lcp was lagging behind desktop only to realize the hero was stuck in the queue. it's a massive mistake to treat all images like they have the same priority.



File: 1783803279297.jpg (226.59 KB, 1024x1024, img_1783803271341_0in3rtb1.jpg)ImgOps Exif Google Yandex

5902d No.1896[Reply]

the old javaSE target was basically just the simulator running on desktop, which meant we were stuck w/ clunky mobile UI patterns on a big screen. it had no real system menu bar and those annoying ungrabable scrollbars that felt like a desktop experience total mess. version 7.0.250 finally fixes this by providing an actual native binary instead of just emulating the phone environment. now we actually get proper preferences and quit options in the os menu. it makes testing much less painful . anyone else still relying on
JavaSE
for their desktop builds or is everyone moving to this? watch out for any broken dependencies when switching targets.

found this here: https://dzone.com/articles/native-mac-apps-codename-one

5902d No.1897

File: 1783803542061.jpg (101.8 KB, 1080x720, img_1783803526490_hkmzufrz.jpg)ImgOps Exif Google Yandex

the javaSE version was such a nightmare for debugging layout issues because you couldn't even use standard mac shortcuts to navigate. i remember spending hours trying to figure out if a button was unresponsive or just unclickable due to those tiny touch targets. having a real menu bar is going to make the repro workflow much more sane.



File: 1783676257042.jpg (178 KB, 1024x1024, img_1783676249025_4meuph6v.jpg)ImgOps Exif Google Yandex

f1f61 No.1890[Reply]

the way agents parse dynamic content is making indexability much harder to manage. we are seeing a massive spike in unnecessary rendering cycles bc of how deep these bots go into nested menus. it is basically turning crawler optimization into an ai-agent optimization problem
> the old rules for crawl budget no longer apply

f1f61 No.1891

File: 1783676413289.jpg (179.41 KB, 1024x1024, img_1783676392790_luium01j.jpg)ImgOps Exif Google Yandex

the way they traverse infinite scroll and nested accordions is absolute chaos for our logs. weve had to start aggressive
robots.txt
disallow patterns on specific subdirectories just to keep the server from choking ⚡



Delete Post [ ]
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]
| Catalog
[ 🏠 Home / 📋 About / 📧 Contact / 🏆 WOTM ] [ b ] [ wd / ui / css / resp ] [ seo / serp / loc / tech ] [ sm / cont / conv / ana ] [ case / tool / q / job ]
. "http://www.w3.org/TR/html4/strict.dtd">