Ask HN: What useful AI tools do you use every day?
33 rajkumarsekar 56 6/25/2025, 4:53:15 AM
There are thousands of AI tools launching every month, but very few become part of our daily workflow.
I’m curious, what AI tools or features do you genuinely rely on every day? This could be anything from coding copilots and writing assistants to niche productivity tools, automations, or personal hacks using LLMs.
Google Lens to identify plants.
BirdNet is Shazam for birds.
ChatGPT for vague questions about everything. I feel like a child asking mom about the world again. It reduced the friction of curiosity.
It’s pretty wild that I have a real world Pokedex.
No comments yet
I’m using SuperWhisper for voice dictation and casually chatting with Claude Code, which I just start in the Obsidian vault directory.
I’m currently onboarding into a new project and need to gather and structure information quickly. So I just jot down whatever comes to mind — in natural language, no structure.
Claude takes care of organizing it: generating #tags, creating [[links]], and making things retrievable later.
After a couple of weeks, I can ask things like:
• What did I do yesterday?
• Who is “John Doe”?
• Which project uses Java version 21?
• …
Happy to share more if anyone’s curious.
(English reviewed with a bit of help from ChatGPT — non-native speaker here.)
I have a Boox tablet I use to take handwritten notes on any relevant topics during the day. And with that I create the following workflow: 1. transform handwritten notes into a markdown document using chatgpt 2. with this document I ask chatgpt to create a summary of yesterdays activities and a list of unfinished things I still need to work on today.
This approach isn't perfect but is surprisingly useful.
2. Claude I use Claude projects for writing articles and for SEO optimisation
3. Cursor I use Cursor for coding daily. Now quite used to the flow it creates and makes my coding super fast
4. Sora and Gemini For image generation. Mostly I need that to share social media posts
All other AI tools come and go but these are 4 constants for me for last few months.
I am often using a mobile device and away from my computer, but want to put something in my Obsidian notes or a Trello card.
ChatGPT speech rec is excellent (so much better than iOS), so often I will start a temporary chat, press the Mic button, and say something like :
“Take the following dictation and repeat it back to me without adding anything. Don’t alter the content, but clean it up if needed… “. Then I ramble on for a while, with pauses when I need to gather my thoughts for the next bit. Then, when I get the paragraph back, I long press for the copy option and paste it back into my other app.
No comments yet
But I think your question is fundamentally flawed—-we all use AI in phone and editor autocomplete, searching, summarization and whatnot on a daily basis now in office tools. It is using it for actual useful output that counts, and for that it is still below what I deem acceptable.
- Mistral as search engine and for natural language processing (mostly summaries of something)
- Co-Pilot, mostly as autocomplete and sometimes to ask questions or let it write tests or do refactoring
- AWS AI Services to create tasks, tickets and stories from templates in JIRA or YouTrack.
- Github/Co-Pilot and Gitlab/snyk for code review and security analysis in PRs
[1] basically https://github.com/roskakori/wolkenlose-ki-fuer-zu-hause/tre... and now i try also https://github.com/ggozad/haiku.rag
Granola - transcription and meeting notes, searching across notes, recalling action items
I've played around extensively with ChatGPT, but Perplexity now covers my use cases. I'm looking to test Claude, primarily because Perplexity does not currently support MCP servers, and I need an assistant who can answer questions across all my work files (Google Drives, Calendar, Slack messages, GitHub, etc.).
[1] https://www.deepl.com/en/translator
wow something that you can find literally in the second sentence of the wikipedia article for haydari.
Btw, it's the third sentence.
it is in the second sentence.
https://en.wikipedia.org/wiki/Haydari
The saying "Can't see the forest for the trees" really makes a lot of sense now.
Self-marketing: Started Marvai this week https://github.com/StephanSchmidt/marvai/ as an AI tool for installing useful Claude Code prompts
Very rarely I need a bash script or systemd service written from a command line, or just something where I know what to search for and what to replace it with.
Then I use Co-Pilot.
The Jetbrains code helper AI is 99% useless, also inconsistent.
i find it helpful to sort of help me make sense of things when i have lots of ideas and unable to structure it
I've even used it for emergencies - trying to locate a place while visiting another city and both maps and the place's insta guide failed .. it was pouring down there. (maybe that's why im partial lol)
i have also started enjoying notebooklm for contextual content generation
gemini - ive tried but it keeps disappointing. currently exploring perplexity
I may use a tool time to time, but I do not use anything daily (on purpose at least, as all software nowadays has AI running in the background that I do not care for).
1. notebooklm for deep-dive into any document
2. Notion AI for QA on my own documents (works really well)
3. cartesia.ai for very good and cheap audio generation
4. veed.io for automated shorts generation with voiceovers and background imagery
5. zenquery.app for data analysis on my large csv and json files
6. regrowth.so for building my own brand on twitter by copying others
7. syft.ai for news summaries (actually works)
Then I have it give me:
- A transliteration (I'm still getting used to the non-Latin alphabet)
- A list of vocabulary from the story
- Grammar tips
Then it is emailed every day around lunch.
I need to go back and tweak it, though, because Gemini really likes starting stories off saying that the sky is clear and the sun is shining.
So what I do is I take notes in a doc throughout the week (Obsidian periodic weekly notes) and just send all of it to a custom GPT that creates the update for me. I usually then spend ~10 minutes cleaning it up.
I'm generally allergic to having an LLM write for me, but seeing as it _must_ be in the requested template and ChatGPT saves me a couple of hours, I let this case slide.
Local LLM: deepseek r1 and qwen2.5coder; phi4 for my project.
Pycharm pro as my main IDE. I <3 pycharm.
I have been giving Void IDE a try this week but it's just frustrating me more than anything.
Has to be a bug where when i approve it to edit the file and it reads maybe 5 lines, stops, still says its running but isnt doing anything. then the chat keeps going like an error happened; but then it really errors out. I'm not ruling out that im just screwing it up somewhere.
Duckduckgo's llm offering with no sign in required and more privacy.
Warp because I can't remember many commands (copilot in the terminal works too but you lose agent mode). They've just release v2 today, looking forwards to try it as a free Claude Code.
Looking to switch to Dia or some agentic browser as main driver but at the moment content with Firefox and Grok as the default search engine.