Junho Yeo (@_junhoyeo) header
Junho Yeo (@_junhoyeo) profile picture

token accountant @tokscale @hey_arca

Joined September 2020
Photo by Junho Yeo (@_junhoyeo)Photo by Junho Yeo (@_junhoyeo)Photo by Junho Yeo (@_junhoyeo)Photo by Junho Yeo (@_junhoyeo)Photo by Junho Yeo (@_junhoyeo)Photo by Junho Yeo (@_junhoyeo)Photo by Junho Yeo (@_junhoyeo)Photo by Junho Yeo (@_junhoyeo)Photo by Junho Yeo (@_junhoyeo)Photo by Junho Yeo (@_junhoyeo)
Junho Yeo retweeted
Had a wonderful time in @ycombinator startup school event in Taipei! felt so much energy from the speakers and the people here, they were all proud that they are changing the world and living in an era where one can actually do it. I flew all the way to Taipei to tap @garrytan’s phone with my ring and show what I’m building, so one mission completed, with so much more to go He tapped the ring and connected with me thru it, so my Arca now remembers when and where we first met, and what conversations we had. We are building in stealth, and a gamechanger in AI & context field is coming to the world. I am in delusional belief that this will work, and a lot of people at the event actually wanted to buy one on the spot. If yall are interested in becoming the first 100 friends to receive the product, comment / DM me! And i’ll lyk
1
1
12
376
The new tokscale.ai is live. → Leaderboards for humans, models and agents → A models dashboard: usage, benchmarks, price → Capital: AI funding, compute deals and capex, every number sourced And it loads in a blink now.
5
2
2
12
844
tokscale.ai just got up to 72× faster: ✓ home page, first byte: 3.6 s → 0.08 s ✓ profile page HTML: 16.1 MB → 1.0 MB ✓ tokscale submit, p75: 38 s → 0.6 s ✓ leaderboard, Lighthouse mobile: 67 → 96
1
116
Junho Yeo retweeted
AI coding-agent costs get messy when every tool keeps its own local history. Tokscale is a command-line tracker and visualization dashboard for token usage and costs across AI coding agents. It helps you see what your agents consume by scanning local session data, then organizing usage by model, client, workspace, session, and time. Key features: • Multi-agent scanning – reads local usage from Claude Code, Codex CLI, OpenCode, Gemini CLI, Hermes Agent, and other supported tools • Token-level breakdowns – tracks input, output, cache read/write, and reasoning tokens • Interactive terminal views – browse usage, models, daily, monthly, sessions, projects, stats, and agents in its TUI • Flexible reporting – filter by platform, date range, or year and group usage by model, client, workspace, or session • JSON export – output reports and contribution-graph data for scripting or external visualization It’s open-source (MIT license). Link in the reply 👇
9
7
1
25
2,565
Junho Yeo retweeted
an announcement from the pierre computer company
382
138
161
4,010
1,191,198
how i prompt my agents now
201
990
156
15,538
605,573
Junho Yeo retweeted
We just refreshed every screenshot in the Tokscale README with tuishot, all in one go. If you're building TUIs, this is a real win.
I built tuishot so your coding agent can see the terminal UI it just built. Run the app, press keys, capture the screen. PNG, SVG, text & JSON from the same frame. npx @tuishot/cli -o shot.png -- your-tui
1
3
91
I built tuishot so your coding agent can see the terminal UI it just built. Run the app, press keys, capture the screen. PNG, SVG, text & JSON from the same frame. npx @tuishot/cli -o shot.png -- your-tui
8
1
1
15
715
this post seemed to get a lot of traction, some of which was people asking me to open source my setup for it. although @benjitaylor's incredible skill is a great base, my lab is just a place for me to build components. It's opinionated. I use the components models build to help define my own taste. I added video exporting in case you want to share here. But this is a blank canvas, you make the art. So here it is: byolab (build your own lab) github.com/jal-co/byolab
I HIGHLY recommend anyone who works on interfaces has a personal "lab" or scratch pad they go to for creating components or tuning interactions. use it as a place to curate your own personal library, i regularly revisit this. using Benji's work as a base (example output using something i built earlier):
4
7
1
182
13,312
Junho Yeo retweeted
We just killed Higgsfield Genjutsu. Introducing Sume Remix : same workflow as Genjutsu, 2.6x cheaper > clone any video > no markups, no subscriptions > start from $5 Try now at sume.com/remix
108
85
37
1,098
240,476
Junho Yeo retweeted
이거 진짜 엄청 시끄러워야 될 이슈인데 국내는 조용하네. 몇몇 중소형 언론사가 두드리고 있는 거 같은데 대형 언론사가 쥐잡아먹은듯 조용하니 이슈화가 안됨. 🤣
为什么你花 2 万块买的旗舰智能电视,按遥控器还是卡得像十年前的安卓机? 昨天安全研究员 @log4j 把自家 3000 刀的三星电视 Root 了,扒出来的底层代码堪称年度草台班子大赏: 1. 电视后台每 500 毫秒截一次屏(连 HDMI 投屏看电脑桌面都截),打包发给近 400 家合作商。如果电视没麦克风,后台进程甚至会顺着蓝牙和回音壁去「借」一个麦克风继续听。 2. 最搞笑的来了:为什么菜单这么卡? 另一位研究员 @mike64_t 拆包发现,三星在电视那颗羸弱的 CPU 上,塞了一个叫 alex.tflite 的模型——246 万参数,2012 年古董级的 AlexNet。 这玩意每隔 2 秒抓一张画面跑推理,猜你在玩什么游戏。结果因为颜色通道写反了,它 99% 的时间什么都认不出来,纯烧 CPU。 3. 那如果你在设置里关掉「隐私追踪」呢? 代码逻辑是:先死命跑完 CPU 推理,准备上传时再看开关——发现你关了,就把结果扔进垃圾桶,2 秒后继续跑。 4. 结局更是重量级: 当号主用 OpenAI 逆向分析「自己花 3000 刀买回家的电视固件」时,OpenAI 反手一个 Cyber Abuse 把他号封了,顺带没收了剩下的 500 刀余额。 三星负责偷窥客厅,2012 年的 AI 负责烧干电视 CPU,OpenAI 负责保护三星的屎山代码不被你发现。
9
61
1
224
31,550
Junho Yeo retweeted
We just killed Higgsfield. You can use Higgsfield's Dashboard, MCP, API for 50% price. Higgsfield : $0.46/s for Seedance 2.5 720p Sume : $0.23/s for Seedance 2.5 720p
As a builder in the AI video space, I was terrified about how much markup API providers charge for simply routing the models. Higgsfield API adds a 40% markup even with its maximum advertised discount. Fal adds 100%. Runway adds 414% (Seedance 2 480p) I might be hated by the industry, but : 
 Today, we’re launching Sume API, the first image & video model provider with 0% markup. For example with Seedance 2.5 1080p • 50% cheaper than fal • 50% cheaper than higgsfield api (29% even with biggest discount) • 28.02% cheaper than kie ai The routing infrastructure layer is never the moat. We believe that APIs should be commoditized, and the moats will be built at the application layer. Check the pricing comparisons below, and see how the broken industry is saved.
6
2
33
1,384
Junho Yeo retweeted
As a builder in the AI video space, I was terrified about how much markup API providers charge for simply routing the models. Higgsfield API adds a 40% markup even with its maximum advertised discount. Fal adds 100%. Runway adds 414% (Seedance 2 480p) I might be hated by the industry, but : 
 Today, we’re launching Sume API, the first image & video model provider with 0% markup. For example with Seedance 2.5 1080p • 50% cheaper than fal • 50% cheaper than higgsfield api (29% even with biggest discount) • 28.02% cheaper than kie ai The routing infrastructure layer is never the moat. We believe that APIs should be commoditized, and the moats will be built at the application layer. Check the pricing comparisons below, and see how the broken industry is saved.
466
393
155
5,769
874,677
Junho Yeo retweeted
3D Kurimanju 🍺
7
1,179
40
4,207
153,855
I just built Computer Use that’s cheaper, faster, and acts like a human 🖥️ Powered by @typesafeai’s JEV 1.14, using the new Image/Box feature. It looks at the screen, decides what to do, then clicks and types like a real person. ⚡ Way cheaper than using an LLM 🧑 Acts like a human: sees the screen, picks an action, clicks or types it 📋 Full report of every step: the screenshot, the action it picked and why Demo: I told it to open Wikipedia, search for an article about aliens, copy the link, and post a tweet with the link and some text. - Cost: $0.0019 - Tokens: ~50,000 OPEN SOURCING SOON on my github 👀 (as always)
73
92
7
1,393
75,897
Junho Yeo retweeted
we hired an abg for @tokscale this is what she posted on day one
1
3
10
341
Junho Yeo retweeted
Tokscale. For the blue hour.
You opened your laptop for one small fix. Now the sky outside is turning blue. Photographers set alarms for the blue hour. You just never went to bed. The French have an older name for it: l’heure entre chien et loup, the hour between dog and wolf. The phrase is more than seven hundred years old. It names the moment when your own dog and a wolf start to look the same. (So do a good diff and a bad one.) If you code with agents, you know how these nights go. You try one approach, then another. You switch models after the first one fixes a failing test by deleting it. You rerun a prompt you were sure would work the first time. By morning, the commit keeps only the part that worked. @Tokscale counts all of it. It’s an open-source tool that tracks your token usage and what it would cost, model by model, across more than 50 AI clients. Claude Code, Codex, Cursor, and OpenCode are on the list. So are Hermes Agent, Prime Agent, and Aside. You can turn your usage into a public profile and see where you stand on the global leaderboard. Our name borrows from the Kardashev scale, which ranks civilizations by how much energy they can harness. For anyone building with agents, that energy is tokens. Run npx tokscale@latest to see your own numbers. Next time the sky turns blue, you’ll know what the night cost.
2
4
176
You opened your laptop for one small fix. Now the sky outside is turning blue. Photographers set alarms for the blue hour. You just never went to bed. The French have an older name for it: l’heure entre chien et loup, the hour between dog and wolf. The phrase is more than seven hundred years old. It names the moment when your own dog and a wolf start to look the same. (So do a good diff and a bad one.) If you code with agents, you know how these nights go. You try one approach, then another. You switch models after the first one fixes a failing test by deleting it. You rerun a prompt you were sure would work the first time. By morning, the commit keeps only the part that worked. @Tokscale counts all of it. It’s an open-source tool that tracks your token usage and what it would cost, model by model, across more than 50 AI clients. Claude Code, Codex, Cursor, and OpenCode are on the list. So are Hermes Agent, Prime Agent, and Aside. You can turn your usage into a public profile and see where you stand on the global leaderboard. Our name borrows from the Kardashev scale, which ranks civilizations by how much energy they can harness. For anyone building with agents, that energy is tokens. Run npx tokscale@latest to see your own numbers. Next time the sky turns blue, you’ll know what the night cost.
4
3
1
14
5,997