1,250 results for Token · 2.907s

News for “Token”
20 results • 2883 ms server time
Moozonian News
github.com• Aug 12, 2026• 1 min read
Show HN: Decant – Understand how you spend tokensHey HN - there are lots of tools to understand how many tokens you use and how much it costs, but we haven't found any that tell you where those tokens are going!Decant helps you understand what you are spending tokens on (context gathering, planning, code, chat, etc), so you can optimize it.
Advertisement
Moozonian News
picklebrowser.com• Aug 11, 2026• 1 min read
Show HN: Pickle – token efficient AI agent browser with policy-gated actionsI use agents for a lot of my own work, but whenever they access my browser they end up wasting a lot of tokens reading HTML, so I tried building a browser that simplifies webpage content for them.Pickle is an agent browser and it has all the features a regular browser has (tabs, search history, bookmarks, etc.), but pages load as compact structured data instead of raw HTML to reduce token usage.Overview for those curious:It's policy-gated (blocked domains, actions that need approval like purchases) and every action is logged, so you're always able to see what your agent is doing and can take over at any time.It's also compatible with weaker/local models as it auto-routes by model strength. i.e. small local models are limited to picking actions one step at a time so they don't hallucinate actions or make up element IDs. Stronger models can plan multiple steps ahead.On the token compression side, I've been seeing around 32x less token usage.Other useful features: - shared notebook that y
Moozonian News
overlay.studio• Aug 11, 2026• 1 min read
Show HN: Overlay – Design with your React components and CSS tokens in your IDEHey HN, I'm Francesco. I've spent the last year building Overlay, a UI editor that runs inside your IDE and renders your actual React components (not a redraw of them), using your real CSS tokens.I'm a design engineer and, especially with AI, I found myself spending more and more time in the codebase rather than in Figma. Trying out ideas in code with AI is a lot of fun, but there are specific moments where I need to sketch things out on a classical visual canvas. The problem is, when you work mostly with code and your codebase moves faster than your Figma library, stopping to import your latest components into Figma is an energy-killer.So I built Overlay to get quick access to my React components on a canvas. It's a VS Code extension, so opening the canvas takes seconds. Your components are already there, and you're designing with the real thing: the editor is a normal DOM document, not a WebGL canvas. From there, working with AI (via MCP) is also a breeze. The agent already has acces
Moozonian News
tokenmaxxer.xyz• Aug 3, 2026• 1 min read
Show HN: TokenMaxxer – track every AI token you spend across your coding toolsI use Claude Code, Codex and Cursor (and sometimes Antigravity) basically every day, and could never tell how much I was actually consuming across all of them.So I built TokenMaxxer. A small CLI reads the files these tools already write locally and puts it all in one dashboard, broken out by tool, model, provider and day. It covers 18 tools now, and you get a profile page with your daily activity, cost estimates, and your top models and tools. There's also a global leaderboard if you want to compete against other TokenMaxxers!I'd love to see if anyone can beat the first place (currently sitting at 45.8B tokens).Thanks for stopping by! I'm looking for ways to improve it, so feedback or ideas are much appreciated.
Advertisement
Moozonian News
github.com• Jul 30, 2026• 1 min read
Show HN: Burnless makes LLM context a protocol. Token savings are a side effectI’ve created Burnless because I realized that we were using the context window for things that didn't need to be there anymore.I wanted to stop burning tokens with what had already happened and were no longer necessary for the next step.It was not a discovery that came later. Burnless was born to separate things. What I didn’t know yet was the size of the problem, the size of the gain and how different it was from the path that everyone else was following. What I see nowadays is that everyone wants the context window to be bigger.That helps, of course. But in my opinion, the concept is wrong. A larger window will only allow you to put more things inside it, but it does not solve the fact that we are using the same place for the conversation, memory, decisions, execution, open questions, work history and all the old records of what has already been done.Burnless separates that into layers.Conversations, thoughts, debates, questions, and the history of the work are indexed and stored in
Moozonian News
news.ycombinator.com• Jul 28, 2026• 1 min read
SOTA on the hardest AI memory benchmark (BEAM, 10M tokens), with a smaller modelHey HN. I'm Johnny, founder of Exabase. We just hit the highest reported score on BEAM, the hardest AI memory benchmark, at every scale up to 10 million tokens. We also ran our evaluation using Gemini 3 Flash, when all previous leaders depended on a much larger model (Gemini 3 Pro).At 10M tokens, the scale is vastly larger than any model's context window, so context stuffing isn't an option (aside from the fact that only about half of a large window can be effectively utilised without degradation). The only way to score well is recall that fundamentally works.Our system (M-1) scored 76.9% at 100K, 75.0% at 1M, and 68.0% at 10M. Previous leaders were Hindsight (73.4%, 73.9%, 64.1%) and Honcho (63.0%, 63.1%, 40.6%), both using Gemini 3 Pro, while we used Flash.We saw the competitive gap get wider at scale: 3.5 points ahead of Hindsight at 100K, 3.9 at 10M. The gap with Honcho goes from 13.9 to 27.4 points. As the corpus gets bigger, it filters out effective recall vs. brute-forcing / mod
Moozonian News
ai-meter.app• Jul 25, 2026• 1 min read
Show HN: AI Meter – Local token usage with energy and water estimates[Problem] I wanted to track how many tokens I am burning and it's estimated energy and water impact.[WHAT THIS IS] A local token usage tracker that works across different coding harnesses and estimates energy and water impact[MY EXPERIENCE] Tracking this really put things into perspective. My personal usage is close to 1 MWh in the past three months which I had no idea about. This does not include my AI usage at work which is probably 10x.[DEMO] If you are interested in tracking these numbers locally, try out: https://ai-meter.app- It reads historical data if the files are present. Works completely locally (except to check if there is an update for the widget) - No login. No analytics. No tracking.----[On estimates]AI Meter measures tokens from local AI coding-tool logs, then estimates:Electricity: tokens ÷ 1M × 0.39 kWh, calibrated from Oviedo et al.’s production-scale inference model.Direct cooling water: electricity ÷ 1.20 PUE × 0.45 L/IT-kWh, using LBNL’s U.S. data-center WUE scena
Advertisement
Moozonian News
meetmaxx.co• Jul 17, 2026• 1 min read
Show HN: Maxx – Live token tracker in the cli (Claude)Token management is a big deal when you have client projects to deliver. I need to have a pulse on what my current burn looks like, especially because I am spawning a bunch of subagents.Agents need to know in real-time what the current utilization is. I went through 10 iterations of this tool - and likely going to do a number of more iterations. The net net is that I needed something to show me if I am on pace to use the tokens alloted to my subscription.Using up the session (maxxing out my session) is not the right way because you run out of tokens by Wednesday at 6pm roughly. So how do you tell your agents that are running in the middle of the night to "not use too many tokens"?So the solution I found was a 5hr ROLLING window. Tokens are allocated across the 7 day period as if they were in chunks. Say 100k tokens an hour for easy math.500k tokens per session, 10 sessions left in the period. 5M tokens remain.Now - the software updates each second. Based on your usage in the window, yo