Alternatives
Products that do what ISON – Data format that uses 30-70% fewer tokens than JSON for LLMs does
ISON (Interchange Simple Object Notation) - a data format optimized for LLMs and Agentic AI. The problem: JSON wastes tokens. Curly braces, quotes, colons, commas - all eat into your context window. ISON uses tabular patterns that LLMs already understand from training data: JSON (87 tokens): { "users": [ {"id": 1, "name": "Alice", "email": "[email protected]"}, {"id": 2, "name": "Bob", "email": "[email protected]"} ] } ISON (34 tokens): table.users id:int name:string email 1 Alice [email protected] 2 Bob [email protected] Features: - 30-70% token reduction - Type annotations - References between…
- 1JA
2019 · github.com
- 2
- 3
- 4IA
2016 · github.com
- 5
Stop wasting money on JSON. Save 50%+ tokens on your LLMs
18d ago · zeon-eight.vercel.app
- 6AT
I've been exploring how to describe UI layouts to LLMs efficiently. The problem: When you ask an AI to generate or modify UI, how do you describe the current state? - Natural language ("header on top, form below") is ambiguous - ASCII art breaks when edited (alignment issues) - HTML is precise but verbose I ran some measurements. For a simple login form: - Natural language: 102 tokens - ASCII art: 84 tokens - HTML: 330 tokens I experimented with a grid-based text format using Excel-like cell references: grid: 4x3 A1..D1: { type: txt, value: "Login" } A2..D2: { type: input, label: "Email" }…
Feb 2026 · github.com
- 7JM
2012 · github.com
- 8
- 9LL
2014 · github.com
- 10TA
TYON (typed object notation) is my attempt to address the main frustrations I have with JSON. 1. Key repetition - you can specify the key names once at the start of a list or map, instead of on each object separately. 2. Escaping strings - everything in a string is literal except for " which is escaped as "". This makes writing regexes and multiline text easier without introducing a bunch of different string styles or excess escaping, which I am not a fan of. 3. While not nearly as important as the first two to me, I did reduce symbol clutter when possible. Files are implicitly maps, keys…
2023 · github.com
- 11CL
2018 · github.com
- 12
- 13QI
2017 · quicktype.io
- 14PR
Hi HN, While building RAG agents, I noticed a lot of token budget was wasted on formatting overhead (HTML tags, JSON structure, whitespace). Existing solutions felt too heavy (often requiring torch/transformers), so I wrote this lightweight, zero-dependency library to solve it. It includes strategies for context packing, PII redaction, and tool output compression. Benchmarks show it can save ~15% of tokens with negligible latency overhead (<0.5ms). Happy to answer any questions!
Dec 2025 · github.com
- 15JB
2015 · github.com
- 16DB
We’ve just launched NUM [1] in the UK, it’s an alternative to the semantic web [2] but based in DNS. We’ve published 23 million data points about 4.8m UK businesses. We had some really interesting HN feedback when we previously posted about one of our example apps [3] so keen to hear what the community has to say this time round. 1. https://www.num.uk 2. https://num.uk/blog/we-crawled-5m-uk-websites-and-published-23m-data-points-to-dns 3. https://news.ycombinator.com/item?id=27598164
2021
- 17MO
2020 · jsonformatter.live
- 18
Token-efficiency linter for LLM prompts and payloads - ritenv/tokensift
8d ago · github.com
- 19OS
2023 · github.com
- 20JA
2019 · github.com
- 21A1
I've seen a lot of comments about how complex frameworks like LangChain can be. Over the holidays, I wanted to see how minimal an LLM framework could get if we stripped away everything non-essential. The result is an LLM framework in just 100 lines of code. These 100 lines capture what I see as the core abstraction of most LLM frameworks: a nested directed graph that breaks down tasks into multiple LLM steps, with branching and recursion to enable agent-like decision-making. From there, you can layer on more advanced features like agents, RAG, task decomposition, and more. I’ve intentionally…
2025 · github.com
- 22JD
We've all been there: you have some JSON data you need to make sense of, so you Google "json formatter" and end up at a site ridden with ads. I used to look at at lot of JSON (mostly log data) for $oldjob, so I set out to build something better. I know there are a lot of tools in this space, but I've put a lot of love into JSON Dive: things like Vim keyboard shortcuts to navigate, dark mode, timestamp/image previews, and multi-file-format-support (XML-in-JSON was a format I dealt with in the context of LLMs/tool calls). I've also made an effort to ensure it can handle large files,…
Sep 2025 · jsondive.app
- 23SS
Hi HN, I’m building SEE (Semantic Entropy Encoding): a searchable compression format for JSON/NDJSON. Goal: reduce the “data tax” (storage/egress) and “CPU tax” (decompress/parse) by keeping JSON searchable while compressed, with page-level random access. I just published a proof-first evaluation release: Offline DEMO ZIP (~10 min): prints compression ratios + skip rates + lookup latency (p50/p95/p99) DD pack: audit/repro evidence (decode mismatch=0, extended mismatch=0, audit PASS) Latest release:…
Feb 2026 · gitlab.com
- 24JT
2019 · github.com
Ranked by how close each launch is in meaning, then by votes. Refine with a description →