There’s a new illness. I call it “LLM Dependency Syndrome.” You can save that $1,200/month by writing simple code: >extract phone numbers: regex >check profanity: blacklist >reformat JSON: json parser >uppercase text: .upper() These few lines of code are faster, cost nearly $0,

For devops/systems ppl, modern terminal emulators on Mac such as Warp or Ghostty are not good when your job involves ssh'ing into multiple VMs. There are UX quirks. They work alright only on the local file system. For ssh, the default Mac Terminal or iTerm2 are good.

Hello OpenRouter team, please provide option to view 1 week usage if possible. openrouter.ai/activity

Hello Abacus.AI Bindu Reddy, is there an equivalent of Gemini Gems or Perplexity Spaces on ChatLLM? I see the DeepAgent would not let me select the model I want. I just need my system prompt + related knowledge + specific model (including expensive ones) saved with a name.

When coding models keep saying "you're absolutely right"...ask "Can we do XYZ? Or is this an overkill? Provide your honest opinion like an alpha male and do not lie" Z.ai GLM 4.5 doing a decent job here.

GLM Coding plans by Z.ai These are literally the best bang for buck coding plans for experimental projects/vibe coding. I do have office provided Copilot with Sonnet,GPT5,etc. So $36 a year GLM Lite is more than enough for me. z.ai/subscribe?cc=f…

fzf - fuzzy find files delta - better file diff with side by side view options z - magical directory navigation dust - better du ouch - single tool for compress/uncompress all formats like zip, gz, 7z, tar fd - better find (skips gitignored files) Kubetail - k8 logs

Put this simple function in your .zshrc or .bashrc When you have the urge to try out a new idea or test a tool/library, you don't want to be bogged down by setup tasks. Put your standard gitignore in $ HOME/templates dir and do this: starter my-app (script in alt-text)

With AI, robotics, humanoids & the coming singularity - will also see a rapid & widespread rise of spirituality. People would get very attuned to human energy on the other side. We call it prana,shakti, qi, pneuma. People would sense this life force regardless of time & space.

Perplexity won’t survive in India with $20/month pricing. They have to reduce price as models become better & cheaper in few months. Current pro users are mostly on Airtel/other tie ups. After 1 year, people would alternate between Gemini/Chatgpt/Grok to deal with rate limits

Kimi / Kimi.ai team has a clumsy way of naming versions. Kimi K2 0905 is named so because of their Sept 5th release. What if the next K2 minor release is on 1st Feb 2026? Will it be Kimi K2 0201 ? How will a user know which is old vs new? Pls be aware of sort order

kubectl apply -f file.yaml will just output a one line info about as to happened. Kubernetes beginners (and troubleshooters) should use verbosity levels (--v=N) when doing kubectl apply to get the behind the scenes log trace/debug details. Eg: kubectl apply -f file.yaml --v=7

A quick bash functions to git add & commit your changes with an LLM generated commit message (ie lazycommit ). Based on github.com/simonw/llm by Simon Willison Install & configure llm. Then put these 3 funcs in your bashrc/zshrc (script in image ALT) lmgflash() gcmessage() gsave()

Developer tip: Always have a “starter” bash function to go from an idea to the IDE Open terminal & type something like: starter my-app This should: * create a new “my-app” dir * copy boiler plate like .gitignore * open VS code/Jetbrains IDE * cd into the new dir on terminal.

Inspirations have expiration date. The idea to IDE gap should be just 5 seconds of you opening terminal & typing “starter my-app”. It should: * Create my-app dir in your workspace. * Copy: gitignore, scripts/prompts, yamls, boilerplates. * Run “<your-fav-ide> my-app”

Too many tabs & content to read? Don't bother if you haven't visited those links for > 8 hours. Just increases confusion(and RAM). Force Quit the browser & start afresh. With AI, the "stuff to read" is infinite. What matters at THIS moment to get work done is all that matters.

Coding agents should just put a ding sound (alert) when they are asking for approvals to run a command and other human in the loop matters.