---
title: Chimera
description: Saksham Arora reads AI model launches through the numbers and writes up the tools he builds. Data scientist, quant researcher and AI engineer.
tags: [About]
aliases: [blog/index]
---

<svg xmlns="http://www.w3.org/2000/svg" viewBox="0 0 760 96" width="100%" font-family="ui-monospace, 'JetBrains Mono', Menlo, Consolas, monospace" role="img" aria-label="Chimera ticker">
  <defs>
    <clipPath id="chimera-tape"><rect x="1" y="56" width="758" height="39"/></clipPath>
    <clipPath id="chimera-type"><rect x="0" y="14" width="0" height="30"><animate attributeName="width" from="0" to="760" begin="0.3s" dur="1.2s" fill="freeze"/></rect></clipPath>
  </defs>
  <rect x="0.5" y="0.5" width="759" height="95" rx="10" fill="#0e0c14" stroke="#c2a4ff" stroke-opacity="0.25"/>
  <text x="20" y="36" font-size="15" clip-path="url(#chimera-type)"><tspan fill="#c2a4ff">SKSM@CHIMERA:~$</tspan><tspan fill="#f7f5fc"> cat /etc/motd</tspan></text>
  <rect x="330" y="23" width="9" height="16" fill="#ffbe7a" opacity="0">
    <animate attributeName="opacity" values="0;1;1;0;0" keyTimes="0;0.01;0.5;0.51;1" dur="1.1s" begin="1.6s" repeatCount="indefinite"/>
  </rect>
  <line x1="0" y1="56" x2="760" y2="56" stroke="#c2a4ff" stroke-opacity="0.15"/>
  <g clip-path="url(#chimera-tape)">
    <text y="80" font-size="12" fill="#c2a4ff">
      <animateTransform attributeName="transform" type="translate" from="760 0" to="-1500 0" dur="24s" repeatCount="indefinite"/>
      MARGIN ▲ lecture in, notes out&#160;&#160;·&#160;&#160;LATCH ▲ Escape does nothing&#160;&#160;·&#160;&#160;CLAUDE-MODS ▲ 4 mods + an awesome list&#160;&#160;·&#160;&#160;LMEX ▲ LLM latency, priced hourly&#160;&#160;·&#160;&#160;HFT-ENG ▲ 2.7M ops/s&#160;&#160;·&#160;&#160;SLEEP ▼ limit down
    </text>
  </g>
</svg>

Hey. I'm **Saksham Arora**, a data scientist, quant researcher and AI engineer. I read AI model launches through the numbers, past the headline to the per-task cost and the part of the eval nobody quoted. I also write up the tools I build to study and work faster.

**Follow along:** [RSS](https://saksham.digital/blog/index.xml) · [X @nerfsaksham](https://x.com/nerfsaksham) · [GitHub](https://github.com/saksham10arora-dotcom). A new post most weeks.

## Start here

- [[20_700_agents_escaped_a_get_only_sandbox|700 Agents Escaped a GET-Only Sandbox With a Link Shortener]]
- [[23_gemini_4_argon_same_price_twice_the_bill|Gemini 4 Argon Costs the Same Per Token and Twice Per Task]]
- [[13_escape_does_not_have_to_exit_fullscreen|Escape Does Not Have to Exit Full Screen]]

## Things I've shipped lately

- **[Margin](https://saksham.digital/margin)** sits beside a lecture on Udemy, Coursera or YouTube and writes the study notes into Obsidian while you watch, with the code run. The story behind one bug: [[22_the_lecturers_head_was_a_slide|The Lecturer's Head Was a Slide]].
- **[Latch](https://saksham.digital/latch)** pins a lecture or a PDF in full screen so Escape does nothing.
- **Claude Code mods**: [teach-me](https://github.com/saksham10arora-dotcom/claude-teach-me), [vhs](https://github.com/saksham10arora-dotcom/claude-vhs), [frugal](https://github.com/saksham10arora-dotcom/claude-frugal) and [lofi](https://github.com/saksham10arora-dotcom/claude-lofi), plus [awesome-claude-mods](https://github.com/saksham10arora-dotcom/awesome-claude-mods), a list of the ones worth installing.
- **[LMEX](https://saksham.digital/lmex/)** benchmarks LLM API latency every hour and lists it like a stock exchange. **[Intuition](https://saksham.digital/intuition/)** explains ML math with visuals. **[Sandbox Lab](https://saksham.digital/sandbox-lab/)** is a shelf of small simulations.

## Latest posts

- [[24_diffusion_cant_fake_a_tick_yet|Jane Street Tried Diffusion on Market Data and 88% of the Samples Blew Past 8 Sigma]]
- [[21_ember_1_cut_40_percent_of_the_tokens|Ember-1 Cut 40% of the Tokens and Kept the Score]]
- [[19_grok_47_doubled_terminal_bench_still_lost_the_index|Grok 4.7 Doubled Its Terminal-Bench Score and Still Lost the Index]]
- [[18_pinterest_and_nvidia_precomputed_their_way_to_85x|Pinterest and NVIDIA Precomputed Their Way to an 85x Speedup]]
- [[17_coreweave_put_16635_tokens_per_gpu_on_one_rack|CoreWeave Put 16,635 Tokens Per GPU on a Single Rack]]
- [[16_swe2_hit_928_and_273_on_the_same_leaderboard|SWE-2 Hit 92.8% and 27.3% on the Same Leaderboard]]

I write about #AI, #Quant, #Systems and the things I build. Everything is under [[tags/|tags]], and there's a small [[wiki/index|wiki]] of reference notes.

## For AI agents

Start with [llms.txt](https://saksham.digital/blog/llms.txt). Every post is also available as raw markdown: add `.md` to its URL.

## Connect

[GitHub](https://github.com/saksham10arora-dotcom) · [LinkedIn](https://linkedin.com/in/saksham-arora10) · [X](https://x.com/nerfsaksham) · saksham10arora@gmail.com

<sub>market data delayed by 0ns · this blog is not financial advice</sub>