Author profile

Khushighosh

I turn caffeine, curiosity, and crazy ideas into working projects. 🚀 Building for the web, experimenting with AI, and learning by creating.

@khushighoshBangalore, KarnatakaView on DEV

Articles by @khushighosh

Browse the latest writing surfaced through DevArt.

#ai#ollama#opensource#llm

I Ran Local LLMs on a Laptop With No GPU: Here Are the Real Numbers

Most local-LLM posts are written on machines with a good GPU. Mine doesn't have one, so I'm sharing...

Khushighosh2 min read

9 reactions1 commentsRead full article

#devchallenge#hf26challenge#ai#opensource

Undead Agent - My AI Agent Dies, Comes Back, and Tells Me What It Saw

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 4: Ghost in the Machine 👻 I...

Khushighosh6 min read

17 reactions12 commentsRead full article

#architecture#llm#programming#softwareengineering

Durable execution for local LLM workflows: what I’m learning

Most LLM demos assume the process stays alive. Kill it midway and the plan, the tool results, and the...

Khushighosh2 min read

9 reactions4 commentsRead full article

#webdev#javascript#debugging#programming

console.log is a legitimate debugging tool

I've used breakpoints, profilers, and all the fancy devtools. But when something breaks at 1 AM, my...

Khushighosh1 min read

8 reactions0 commentsRead full article

#devchallenge#hf26challenge#gemma#tabpfn

Let It Die: The Lazy Email Agent on TabPFN, Gemma and Temporal

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 3: Slow and Steady ...

Khushighosh6 min read

20 reactions18 commentsRead full article

#ai#beginners#programming#webdev

Using AI coding tools without losing your fundamentals

AI can write a working login route in 10 seconds. That’s amazing, and also a little dangerous. If the...

Khushighosh3 min read

7 reactions3 commentsRead full article

#beginners#git#windows#tutorial

6 Windows + Git traps I hit this week (and the 30-second fixes)

I spent this week setting up a public GitHub repo on Windows. Almost every problem came from a tiny...

Khushighosh2 min read

6 reactions0 commentsRead full article

#ai#llm#sideprojects

Three things that surprised me while building this: Re-runs were identical - Same model, same prompt, `temperature: 0`: byte-for-byte the same failures every time. That turned prompt tweaking into real A/B testing. A regex beat the prompt. I

Order Ledger: a WhatsApp-to-ledger side hustle that runs offline on open models ...

Khushighosh1 min read

0 reactions0 commentsRead full article

#devchallenge#hf26challenge#ollama#ai

Order Ledger: a WhatsApp-to-ledger side hustle that runs offline on open models

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 2: Side Hustle ...

Khushighosh5 min read

6 reactions0 commentsRead full article

#devchallenge#hf26challenge#samespot#touchgrass

Go Outside, Take a Photo, Let a Tiny AI Argue About It 🌇

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass I asked a...

Khushighosh4 min read

5 reactions2 commentsRead full article

#devchallenge#weekendchallenge#hf26challenge#studybuddy

Study Buddy: an offline GATE study partner for my friend, built with Ollama

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend What I...

Khushighosh3 min read

5 reactions0 commentsRead full article

Khushighosh

Articles by @khushighosh

I Ran Local LLMs on a Laptop With No GPU: Here Are the Real Numbers

Undead Agent - My AI Agent Dies, Comes Back, and Tells Me What It Saw

Durable execution for local LLM workflows: what I’m learning

console.log is a legitimate debugging tool

Let It Die: The Lazy Email Agent on TabPFN, Gemma and Temporal

Using AI coding tools without losing your fundamentals

6 Windows + Git traps I hit this week (and the 30-second fixes)

Three things that surprised me while building this: Re-runs were identical - Same model, same prompt, `temperature: 0`: byte-for-byte the same failures every time. That turned prompt tweaking into real A/B testing. **A regex beat the prompt.** I

Order Ledger: a WhatsApp-to-ledger side hustle that runs offline on open models

Go Outside, Take a Photo, Let a Tiny AI Argue About It 🌇

Study Buddy: an offline GATE study partner for my friend, built with Ollama

Three things that surprised me while building this: Re-runs were identical - Same model, same prompt, `temperature: 0`: byte-for-byte the same failures every time. That turned prompt tweaking into real A/B testing. A regex beat the prompt. I