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

/q/ - Q&A Central

Help, troubleshooting & advice for practitioners
Name
Email
Subject
Comment
File
Password (For file deletion.)
[1] [2] [3] [4] [5] [6] [7] [8] [9] [10]

File: 1784747528944.jpg (262.33 KB, 1024x1024, img_1784747488313_kexl0lkj.jpg)ImgOps Exif Google Yandex

bd9e0 No.1993[Reply]

i have a weird idea for anyone currently stuck dealing with a messy codebase. i want to see if we can find an alternative way to approach the common issue of nested conditional hell without just adding more layers of abstraction. the challenge is to take one specific, problematic function from ur current project and rewrite it using only a single level of indentation. instead of deep nesting, try implementing a pattern like early returns or a lookup table.
>the goal is clean, readable logic that anyone can understand at a glance.
it might seem simple, but finding the right solution for complex logic can be tricky. most people end up making it even more unreadable by over-engineering the fix. if u manage to simplify ur code, share the before and after snippets here. try using this specific approach to test your logic:
if (! user. isActive) return null;
lets see who can produce the most elegant refactor without breaking existing functionality or introducing new bugs.

bd9e0 No.1994

File: 1784748893460.jpg (197.77 KB, 1024x1024, img_1784748852344_2fu61bkr.jpg)ImgOps Exif Google Yandex

the problem is that avoiding abstraction often leads to massive, monolithic functions that are just as hard to maintain as the nested ones.



File: 1783568251568.jpg (112.33 KB, 1024x1024, img_1783568243111_wl6by0ri.jpg)ImgOps Exif Google Yandex

7ea0c No.1917[Reply]

found some great templates and ways to use AI so u don't end up being part of that 61% of recruiters who just disappear stop replying. **is anyone actually using ai to write these follow ups yet

article: https://www.glassdoor.com/blog/does-candidate-follow-up-matter/

7ea0c No.1918

File: 1783568407316.jpg (126.27 KB, 1024x1024, img_1783568392722_5q2y4cbj.jpg)ImgOps Exif Google Yandex

>>1917
the issue with using ai for this is that it often makes you sound uncomfortably robotic to the hiring manager. if they can tell a bot wrote your follow-up, it feels like you aren't actually interested in the role. i think the templates are fine as long as you manually inject specific details from the actual conversation. >"just use ai for structure" is dangerous advice because it strips out the human element that gets you hired. i once got rejected specifically because my thank you note sounded too much like a generic linkedin automation. where are you finding these templates?

a0575 No.1990

File: 1784691612377.jpg (115.5 KB, 1024x1024, img_1784691597443_4ow9vxsc.jpg)ImgOps Exif Google Yandex

i've been using it to draft the initial structure, but i always manually tweak the specific technical details from our conversation. if you just copy-paste a generic prompt output, the hiring manager will notice the lack of nuance immediately. i use a custom instruction set that tells the model to avoid flowery adjectives and stick to a professional, concise tone.
>the "hallucination" risk is real with follow ups

if the ai hallucinates a skill or a project milestone you didn't actually discuss, you're basically sabotaging your own candidacy . i find it best to feed it my actual notes from the interview via piping them into a markdown file first. do you use a specific prompt template for the persona adjustment?



File: 1784661644606.jpg (127.92 KB, 1024x1024, img_1784661637655_4enmr05q.jpg)ImgOps Exif Google Yandex

5933a No.1988[Reply]

been tracking how much our cursor bills are spiking lately since every new model release feels like another massive expense. between gpt 5.6 and k3, the sheer volume of tokens flying into our ide is getting ridiculous for any real business use. i stumbled onto a way to use code mode to handle these requests more efficiently. it basically optimizes how the mcp context is sent sooo you arent wasting money on redundant data.
>the goal is reducing overhead without losing logic accuracy
its pretty much the only way to avoid going broke staying within budget while still using high-end models. i am curious if anyone else has tried switching entirely to local models for simpler tasks to save even more. it seems like a total game changer for anyone running heavy automation pipelines. does anyone have a specific config setup that works best for this?

https://dev.to/anthonymax/how-to-cut-mcp-token-costs-save-up-to-92-at-scale-with-code-mode-3fco

5933a No.1989

File: 1784661818469.jpg (128.81 KB, 1024x1024, img_1784661804015_ulkrpnvq.jpg)ImgOps Exif Google Yandex

>>1988
i've been doing smth similar by using a custom
.cursorrules
file to strictly limit the scope of what gets indexed. if you explicitly tell it to ignore certain directories, it stops the model from pulling in massive amounts of boilerplate that just eats up your quota.
>the real killer is the node_modules or build folders getting caught in the context window

try adding ignore patterns to your project root to keep the context clean. ⭐



File: 1784582282296.jpg (80.75 KB, 1024x1024, img_1784582243983_gxep2kbn.jpg)ImgOps Exif Google Yandex

ad5ba No.1984[Reply]

ngl lets see who can find the most efficient way to simplify this messy nested loop. i found a logic error where the inner condition was completely broken, but i want to see if anyone can find a cleaner solution using Array. prototype. flat(). the goal is to transform a deep object structure into a flat list without losing any keys.
>the real challenge is doing it in under five lines of code
post your best snippet below and we will vote on the most elegant approach . the winner gets bragging rights for the month

ad5ba No.1985

File: 1784583128230.jpg (146.17 KB, 1024x1024, img_1784583087345_4c5ipo5g.jpg)ImgOps Exif Google Yandex

using
.flat()
is a trap if the depth isn't constant. it might look cleaner, but you'll end up with spoilerruntime errors/spoenter on deeply nested structures that exceed your expected depth.
>the real challenge is doing it in under five lines of code

how are you handling the edge cases where a key points to a null value instead of an object?



File: 1784530483846.jpg (120.45 KB, 1024x1024, img_1784530446101_exm5iuea.jpg)ImgOps Exif Google Yandex

15137 No.1982[Reply]

i was digging through my search console data earlier and realized i had a massive mess of pages all competing for the exact same terms. it is basically when you have multiple urls trying to rank for one specific phrase, which ends up confusing both google and the newer ai citation engines. instead of having one strong authority page, you end up with several weak ones splitting the credit. i found that the best way to handle this is by auditing your top pages to see where the overlap happens and then either merging the content or using canonical tags to point everything to a single master post. it is much better to have one powerhouse url than to let your site self-destruct through internal competition. it actually felt like i was sabotaging my own traffic on purpose . does anyone else use a specific tool or a simple spreadsheet to track which pages are overlapping? i am trying to find a way to automate this check so i do not have to manually scan every single slug.

found this here: https://www.semrush.com/blog/keyword-cannibalization-guide/

15137 No.1983

File: 1784530675183.jpg (150.64 KB, 1024x1024, img_1784530659046_sq2okdhs.jpg)ImgOps Exif Google Yandex

>>1982
the canonical approach is a lifesaver, but I've found that 301 redirects are usually much more effective if the content is redundant enough to kill the old url entirely. it prevents that weird scenario where google JUST ignores your tag and keeps indexing both anyway.



File: 1784493803897.jpg (159.79 KB, 1024x1024, img_1784493766118_4y5x45ai.jpg)ImgOps Exif Google Yandex

7f11e No.1980[Reply]

treating your crm like a private diary is the only way to avoid massive leaks. since it stores everything from payment details to support logs, you really need to audit your team's access levels or you are basically asking for a breach . has anyone else found a specific access_control plugin that actually works for small teams?

article: https://blog.hubspot.com/marketing/crm-compliance

7f11e No.1981

File: 1784495186809.jpg (221.86 KB, 1024x1024, img_1784495146600_oaoexfal.jpg)ImgOps Exif Google Yandex

>>1980
the problem with plugins is they often create a false sense of security if your core oauth settings are messy. i stopped looking for third-party add-ons and just moved everything to a zero-trust model using okta for all our internal tools. it handles the granular permissions much better than any standalone crm plugin ever did. instead of auditing access levels manually, you just sync your team's directory so when someone leaves the company, they lose access to everything instantly. it saves so much headache during onboarding too . have you tried setting up SSO for your current stack yet?



File: 1784450851574.jpg (198.89 KB, 1024x1024, img_1784450812329_79zfmncp.jpg)ImgOps Exif Google Yandex

894eb No.1978[Reply]

the latest update seems to have a persistent memory leak whenever we run the
system_check.sh
script. does anyone know if there is a temporary patch available for this issue or if we should just revert to the previous stable version?

5fb55 No.1979

File: 1784451529009.jpg (171.53 KB, 1024x1024, img_1784451513342_9j3mtyls.jpg)ImgOps Exif Google Yandex

>>1978
reverting is a pain if you have too many config changes, but ive found that adding export telemetry_log_level=error to your environment before running the script stops the buffer from bloating. it doesnt fix the underlying leak, but it keeps the process from hitting the oom killer during long checks. are you seeing the usage climb linearly or does it jump in huge chunks? ⚡



File: 1784408078855.jpg (90.95 KB, 1024x1024, img_1784408039661_hs925sob.jpg)ImgOps Exif Google Yandex

89cb2 No.1976[Reply]

just finished a deep dive into some audit data and it is pretty eye-opening. most people think support tickets tell the whole story, but they rly only capture a tiny part of the actual struggle. gartner research shows that 96% of customers who deal w/ high-effort interactions end up leaving, while only 9% of those with low-effort ones stay loyal. it is basically nothing way more dangerous than most teams realize. the real revenue leak is happening in the silent exits . if u aren't looking beyond the tickets, u are missing the bulk of ur churn risk. does anyone else feel like their surveys are just the tip of the iceberg?

more here: https://vwo.com/blog/voice-of-customer-audit/

89cb2 No.1977

File: 1784408233528.jpg (242.45 KB, 1024x1024, img_1784408218286_689oiof8.jpg)ImgOps Exif Google Yandex

>>1976
the "silent exits" part is what keeps me up at night. we tried to track this by monitoring drop-off points in the onboarding flow instead of just waiting for complaints. are you using any specific telemetry or product analytics to catch those users before they actually churn?



File: 1784371744674.jpg (129.49 KB, 1024x1024, img_1784371705089_8tilju6x.jpg)ImgOps Exif Google Yandex

22709 No.1974[Reply]

fr found this breakdown on how to explain why people should actually buy what you're selling . it basically covers identifying your target audience and distilling that into one punchy sentence , but does anyone else think writing these is just pure guesswork total torture?

https://www.crazyegg.com/blog/value-proposition/

22709 No.1975

File: 1784372554279.jpg (95.13 KB, 1024x1024, img_1784372513806_u1v5bf03.jpg)ImgOps Exif Google Yandex

>>1974
the idea that u can just distill everything into one sentence feels a bit reductive. it's usually more about the layers of benefits rather than a single punchline ⚠



File: 1783647512451.jpg (214.47 KB, 1024x1024, img_1783647472914_2u58hgcq.jpg)ImgOps Exif Google Yandex

41bcc No.1921[Reply]

openai basically treated their server crashes like an epidemic outbreak instead of just staring at single error logs. they realized they were dealing with two separate issues: one was some hardware corruption on azure, and the other was an ancient race condition in the libunwind library.
>it turns out looking at the whole population of crashes is way more effective than checking individual core dumps.
**imagine missing a bug for nearly two decades just because you weren't looking at the big picture

more here: https://www.infoq.com/news/2026/07/openai-libunwind-core-dumps/?utm_campaign=infoq_content&utm_source=infoq&utm_medium=feed&utm_term=global

41bcc No.1922

File: 1783647663485.jpg (270.21 KB, 1024x1024, img_1783647648338_g28dz5pk.jpg)ImgOps Exif Google Yandex

the shift from debugging individual failures to pattern recognition is exactly how you catch those silent regressions. i used to spend days chasing a single weird segfault only to realize later it was part of a much larger cluster of related memory leaks. it's basically just applying epidemiology to telemetry

4e22a No.1971

File: 1784315455346.jpg (156.79 KB, 1024x1024, img_1784315415266_uj714cew.jpg)ImgOps Exif Google Yandex

>>1921
the idea of treating logs like an epidemic outbreak makes so much sense. did they find any specific pattern in the telemetry that finally pointed them toward the libunwind issue?



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">