Skip to content
Welcome to my personal blog.About this blog →
deepakshah.net
JournalAboutRSS feed
⌂All articles
AI & Models⌄View all◇Decider 4B: Decisions Without Generating an Answer✳Building My Own AI Chat Harness
Engineering⌄View all▤PDF Parser Benchmark: Three Tools, Three Annual Reports⌘CubeSandbox vs Docker: Startup Speed, Memory, and What We Learned
Field Notes⌄View all↗Improving Tool Pages for Users—and for SEO

Independent notes, shared openly.

On this page
LatestTopicsAbout
A personal blog

Notes, ideas, and things I learn.

Personal notes and articles.

About this blog ↗
Extracted character counts from three annual reports. LiteParse returned output for all three; pdf-inspector returned only 45,449 characters for FY2025; anydoc reported that FY2025 needed OCR and returned no output.
Featured storyEngineering

PDF Parser Benchmark: Three Tools, Three Annual Reports

A single-run comparison of LiteParse, pdf-inspector, and anydoc on three Antelopus Selan Energy annual reports, with table checks and a silent coverage failure.

d.deepakshah.net·Oct 10, 2026·5 min read
Read story ↗

The latest notes

05 ARTICLES
All articlesAI & ModelsEngineeringField Notes
▤
Engineering/Oct 10, 2026

PDF Parser Benchmark: Three Tools, Three Annual Reports

A single-run comparison of LiteParse, pdf-inspector, and anydoc on three Antelopus Selan Energy annual reports, with table checks and a silent coverage failure.

5 min read·deepakshah.net
↗
◇
AI & Models/Sep 29, 2026

Decider 4B: Decisions Without Generating an Answer

A look at one-pass choice models, their inference costs, the v2 and v2.1 tradeoff, and how I would test one in a real workflow.

6 min read·deepakshah.net
↗
↗
Field Notes/Sep 20, 2026

Improving Tool Pages for Users—and for SEO

How behaviour data helps improve user experience, meaningful engagement, and the value of search traffic—with a practical font-tool example.

5 min read·deepakshah.net
↗
✳
AI & Models/Sep 12, 2026

Building My Own AI Chat Harness

Inside my personal AI workspace: pinned conversations, visible subagents, reusable skills, and ways to continue work across chat, voice, and companion clients.

7 min read·deepakshah.net
↗
⌘
Engineering/Sep 10, 2026

CubeSandbox vs Docker: Startup Speed, Memory, and What We Learned

What same-host tests showed about CubeSandbox and Docker: startup time, shared memory costs, and command completion at 100 and 500 microVMs.

10 min read·deepakshah.net
↗

Follow your curiosity

✳AI & Models ↗2 notes to explore⌘Engineering ↗2 notes to explore✎Field Notes ↗1 notes to explore

A space for my writing.

Articles, ideas, and notes from my personal blog.

About this blog →
© 2026 deepakshah.netFollow along via RSS ↗

On this page

LatestTopicsAbout

Personal notes and articles.

Follow via RSS →

Start typing to search all articles.

Enter open first resultEsc close
Browse articles
⌂All articles
AI & Models⌄View all◇Decider 4B: Decisions Without Generating an Answer✳Building My Own AI Chat Harness
Engineering⌄View all▤PDF Parser Benchmark: Three Tools, Three Annual Reports⌘CubeSandbox vs Docker: Startup Speed, Memory, and What We Learned
Field Notes⌄View all↗Improving Tool Pages for Users—and for SEO
AboutRSS feed