Hacker News new | past | comments | ask | show | jobs | submit login
Show HN: Semantic Macros Text Editor (samrawal.com)
69 points by zora_goron 75 days ago | hide | past | favorite | 19 comments



button "pornify" with prompt "rephrase this text as if it were in a 1970s erotic pulp novel" turned:

You can define your own "macros" - instructions for an LLM (like "summarize this text", "fix the grammar") and add them as buttons for easy access. You can also make one-time commands. Then, highlight the text you want to send to the LLM, and click the button!

into:

In the tantalizing world of seductive prose and forbidden texts, imagine crafting your own alluring "macros"—personalized whispers to your erotic muse, coaxing it into decadent actions such as "entice this passage with brevity" or "caress the syntax with meticulous finesse." Fashion these sultry commands into luscious buttons, waiting at your beck and call. With a mere flick of the wrist, highlight the prose that begs for attention, then surrender it to the willing buttons, igniting the LLM's desire to transform mundane words into tantalizing pleasures.


I had to try that out and the output was better than I could ever expect. "Select the coy text you desire to ensnare" and "tantalizing beck-and-call instrucitons for a large language model that purrs with delight" had me in tears

In the dimly lit confines of your mahogany-laden study, nestled between relics of an age gone by, there exists a silky tome of technological desire known as sam.txt, an alluring plaintext editor with the promise of sating your most audacious cravings with its potent language manipulations. Within your exquisite grasp lies the power to conjure sizzling "macros," those tantalizing beck-and-call instructions for a large language model that purrs with delight at your every beckoning, whether to "summarize this prose" or "mend this foolish syntax." These instructions become glistening buttons laid bare for your indulgent convenience, ready to be pressed by those yearning fingers.

In addition, your whims can be indulged without plan or preamble. Select the coy text you desire to ensnare, offer it to the sultry LLM, and press the button that promises sweet satisfaction.

I invite you to embark on this thrilling escapade! Select the tempting passage above and click 'Formalize' to experience ecstasy beyond the written word.


For fun I tried "Formalizing" the original Ecclesiastes verse that Orwell uses in Politics and the English Language, where he translates it into modern English.

The original text is: "I returned and saw under the sun, that the race is not to the swift, nor the battle to the strong, neither yet bread to the wise, nor yet riches to men of understanding, nor yet favour to men of skill; but time and chance happeneth to them all."

Orwell's parody is: "Objective considerations of contemporary phenomena compel the conclusion that success or failure in competitive activities exhibits no tendency to be commensurate with innate capacity, but that a considerable element of the unpredictable must invariably be taken into account."

Formalize turns out: "Upon reflection, it becomes apparent that success—whether in the form of victory in a race, strength in battle, provision for the wise, wealth for men of understanding, or favor for those with skill—is not guaranteed solely based on one's abilities. Rather, time and chance play a significant role in these matters for all individuals."


Cool! I'm working on a note-taking app with a similar (but not identical) feature based on Excel-type formulas - except, semantic formulas, not mathematical. Chaining AI formulas together (combined with non-AI stuff like "get the contents of this web page / find my notes that mention X") can be really powerful.


This sounds very interesting, any more details you can reveal would be appreciated!


- Markdown based note-taking app

- Notational Velocity-style omnibar

- Wikilinks, backreferences, etc.

- Excel-type formulas: find nodes in documents that meet criteria and transclude them, feed content to AI with instructions and insert output in doc (since this is a live formula, you can edit the instructions and re-run), have conversations with AI in doc with context ...

email is in profile if this sounds interesting to anyone


This is the future of iterative software construction.


Nice and neat tool! I wonder which llm model is running in the backend and is there any way to run it locally/self-hosted?


Made a button to "Make it an ASCII Map!" showing relationships between concepts, had a lot of fun. Nice one!


I really like this! Is the source code available somewhere? I'd love to add/contribute to it.


This looks really nice application of LLM. I’ve made my button :)

What is the business model behind?

How long will editor remember my buttons?


Thanks for trying it out! Honestly, no current idea of business model - this was just something I thought would be helpful in my daily workflow so I built it out.

The buttons are stored to localstorage, as is the editor text - everything stays locally (besides what is sent to the LLM). I'm planning on a simple import/export mechanism for transferring buttons across different browsers/computers!


It looks like the buttons are just stored in local storage. So in theory, they'll hang about until you clear that.


How do we trigger the highlighting function? I selected text but didn't see anything.


If you’re wondering what “semantic macros” means, it’s just llm bullshit. Not worth looking at.


i mean, it is pretty much what I expected it to mean. It's a macro for semantic instructions. I didn't really see any LLM "bullshit", just a way to macro using an LLM...

Did you expect this to be something different?


Not the original commenter but I expected a semantic macro to be something that expands on syntax-aware macros (in languages like Scheme or Rust) into doing things based on the semantics of the code. Not sure exactly what that would entail but I was intrigued. Instead it turned out to be just quick buttons for sending something to an LLM. Between Github's Copilot, ChatGPT and LLM plugins for Obsidian those needs have already been covered for a long time.


This is awesome. The best "prompt engineering" tool I've seen yet. So well designed. There's a lot of useful things I might try with this.

Nicely done.

Nits: I'd add some small indicator of progress; an active state to the button; also I'd ditch the "one time button thing". You can already delete buttons so anyone can do one time buttons already.

Here's my user test: https://news.pub/?try=https://www.youtube.com/embed/qh2mq0KJ...


Hi there, thanks for trying it out and the kind words! These are great suggestions - thank you!




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: