Vishal

Reflecting on life, work, and the beautiful mind.

[Vishal @learnedquietly]

I’ve been heads-down building GAIA .. a free, fully open-source orchestration kernel for AI agents.

The goal is simple: make autonomous agents reliable, secure, and affordable for everyone. I’ve just pushed a major update with improved local LLM support and context optimization.

Kernel: https://t.co/vBOerW05JR
Agents: https://t.co/GKUGLAzFHw

What makes GAIA different? Most frameworks (LangChain, CrewAI) are libraries you code against. GAIA is a standalone kernel — an OS that manages, secures, and orchestrates agents like processes.

Core features:
⚡ 10-Phase Control Loop: A deterministic state machine ensuring no mission is ever “lost” or “hallucinated.”
🛡️ Policy Firewall: CEL-based rules enforcing a strict deny-by-default security model for untrusted agents.
🧩 Protocol Interop: Native support for Google A2A and Anthropic MCP out of the box.
📦 4-Tier State Store: Seamless persistence from L1 (memory) to L4 (SQLite) for long-running tasks.
🔐 Tamper-Proof Audit: Every action is cryptographically signed (HMAC-SHA256) for enterprise-grade accountability.
It’s all on GitHub — explore the code, break it, and help build it. 🤝

#buildinpublic #OpenSource #AI #Golang #SystemDesign

[Vishal @learnedquietly]

Jayanagar easily ranks among the best spots in #Bengaluru for early morning cycling... calm streets, fresh air, perfect vibes

[Vishal @learnedquietly]

@earthcurated

[Vishal @learnedquietly]

@Psychodoctor06 In today’s era, the probability of meeting such a truly empathetic person is almost zero.

[Vishal @learnedquietly]

@dmuthuk I understand your point, and I feel the same. Trump is entitled to think from his country’s perspective, but making racist comments was unnecessary. It reveals how he truly views Indians, and that’s disappointing.

[Vishal @learnedquietly]

Quoting @karpathy’s “LLM Knowledge Bases” 🧠

Imagine scaling this into a global AI knowledge cache: jointly maintained by all major AI providers.

Only agreed‑upon frequent facts get stored + refreshed, then shared across assistants. Huge savings in compute + consistency + $$.

[Andrej Karpathy @karpathy]

LLM Knowledge Bases

Something I'm finding very useful recently: using LLMs to build personal knowledge bases for various topics of research interest. In this way, a large fraction of my recent token throughput is going less into manipulating code, and more into manipulating knowledge (stored as markdown and images). The latest LLMs are quite good at it. So:

Data ingest:
I index source documents (articles, papers, repos, datasets, images, etc.) into a raw/ directory, then I use an LLM to incrementally "compile" a wiki, which is just a collection of .md files in a directory structure. The wiki includes summaries of all the data in raw/, backlinks, and then it categorizes data into concepts, writes articles for them, and links them all. To convert web articles into .md files I like to use the Obsidian Web Clipper extension, and then I also use a hotkey to download all the related images to local so that my LLM can easily reference them.

IDE:
I use Obsidian as the IDE "frontend" where I can view the raw data, the the compiled wiki, and the derived visualizations. Important to note that the LLM writes and maintains all of the data of the wiki, I rarely touch it directly. I've played with a few Obsidian plugins to render and view data in other ways (e.g. Marp for slides).

Q&A:
Where things get interesting is that once your wiki is big enough (e.g. mine on some recent research is ~100 articles and ~400K words), you can ask your LLM agent all kinds of complex questions against the wiki, and it will go off, research the answers, etc. I thought I had to reach for fancy RAG, but the LLM has been pretty good about auto-maintaining index files and brief summaries of all the documents and it reads all the important related data fairly easily at this ~small scale.

Output:
Instead of getting answers in text/terminal, I like to have it render markdown files for me, or slide shows (Marp format), or matplotlib images, all of which I then view again in Obsidian. You can imagine many other visual output formats depending on the query. Often, I end up "filing" the outputs back into the wiki to enhance it for further queries. So my own explorations and queries always "add up" in the knowledge base.

Linting:
I've run some LLM "health checks" over the wiki to e.g. find inconsistent data, impute missing data (with web searchers), find interesting connections for new article candidates, etc., to incrementally clean up the wiki and enhance its overall data integrity. The LLMs are quite good at suggesting further questions to ask and look into.

Extra tools:
I find myself developing additional tools to process the data, e.g. I vibe coded a small and naive search engine over the wiki, which I both use directly (in a web ui), but more often I want to hand it off to an LLM via CLI as a tool for larger queries.

Further explorations:
As the repo grows, the natural desire is to also think about synthetic data generation + finetuning to have your LLM "know" the data in its weights instead of just context windows.

TLDR: raw data from a given number of sources is collected, then compiled by an LLM into a .md wiki, then operated on by various CLIs by the LLM to do Q&A and to incrementally enhance the wiki, and all of it viewable in Obsidian. You rarely ever write or edit the wiki manually, it's the domain of the LLM. I think there is room here for an incredible new product instead of a hacky collection of scripts.

[Vishal @learnedquietly]

@Sona_unplugged She didn’t want to waste time on the road....
Always thinking in terms of efficiency...

[Vishal @learnedquietly]

Next week: Palakkad for 7 days. Excited.

I wanted to drive, wife vetoed and booked train😞

[Vishal @learnedquietly]

5 more joined today.
30 users now.
Zero feedback so far.
Slow grind.

[Vishal @learnedquietly]

Last 2 days: manually reaching out to parents on Reddit.

No ads. No growth hacks. Just direct conversations.

→ 25 people downloaded SmallShifts.

Still early, but a few interesting signals:

• People respond more when the message feels like a conversation, not advice
• Strong opinions trigger replies, but not always the right users
• Clear problem > clever messaging
This phase is less about growth, more about understanding.

Talking to users > guessing.

[Vishal @learnedquietly]

Last 2 days: manually reaching out to parents on Reddit.

No ads. No growth hacks. Just direct conversations.

→ 25 people downloaded SmallShifts.

Still early, but a few interesting signals:

• People respond more when the message feels like a conversation, not advice
• Strong opinions trigger replies, but not always the right users
• Clear problem > clever messaging
This phase is less about growth, more about understanding.

Talking to users > guessing.

[Vishal @learnedquietly]

Hi, wanted to share my experience with Apple App Store submission for my app SmallShifts.

Actual App Store Reviewer Fixes (March 2026)

Timeline:
• Apple Developer enrollment: 3 days (stuck 48h → requested callback → activated in minutes)
• App review: 7 days
→ Total: ~10 days

Also: App binary review and subscription review happen separately.
My binary was approved first. Subscription review came ~10 hours later.

These are real issues Apple rejected my app for — not crashes, not bugs. Mostly UX, compliance, and “reviewer expectations”.

Standard EULA (3.1.2c)
Linked my own Terms page → rejected
Fix: Must use Apple’s Standard EULA for subscriptions
Lesson: Don’t customize legal unless you’re enterprise

Anonymous users still need delete (5.1.1)
Even without login, stored data must be deletable
Fix: Local DB wipe + CloudKit purge for guest users
Lesson: No account ≠ no responsibility

iPad testing (2.4.1)
Even though my app was iPhone-only, reviewer tested it on iPad and it broke in split view
Fix: Handle dynamic layout + resizing properly
Lesson: “iPhone only” does NOT protect you from iPad testing

Subscription clarity (3.1.2)
Price/length/title not clearly grouped → rejected
Fix: Explicit summary block above paywall CTA
Lesson: Don’t make reviewer hunt for pricing

Subscription visibility (3.1.1)
Plan buried in settings → rejected
Fix: Moved to top-level settings
Lesson: Manage = easy as purchase

Non-native settings UI (4.0)
Custom accordion → flagged
Fix: Flat navigation with standard patterns
Lesson: Don’t fight iOS conventions

Disabled buttons = “unresponsive” (2.1)
Login button disabled → rejected
Fix: Always enabled + show validation errors
Lesson: Every tap must respond

Offer code failures during review
Generic “Purchase Failed” → friction
Fix: Handle specific error states clearly
Lesson: Make reviewer testing easy

Apple review isn’t about your code.

It’s about how predictable, transparent, and testable your app feels to them.

[Vishal @learnedquietly]

SmallShifts:Parenting Tracker: Don’t fix your child. Shift your approach. https://t.co/4GHaldlBsU via @producthunt

[Vishal @learnedquietly]

@Sona_unplugged Thank you 😊

[Vishal @learnedquietly]

New themes just dropped in SmallShifts.
Different moods, same intention: show up better for your child, every day.

link: https://t.co/NjWMXWYvx8

#parenting

[Vishal @learnedquietly]

Added these 4 new themes to my app...

[Vishal @learnedquietly]

#parenting

[Vishal @learnedquietly]

Any parents here?

This is how our everyday interactions go:

A small thing turns into a tantrum.
You say no. They push back.
You react. It escalates.
Later you think, “I could’ve handled that better.”
But next time… it plays out the same way again.

Sounds familiar?

Every parent goes through this. I did — and I still do.

Most of the time, we try to change our kids.
But maybe it’s the other way around.
Maybe we, as parents, need to change our approach.

That’s what I built — SmallShifts.

You log moments, and over time it shows patterns and suggests small shifts in how to handle them better next time.

https://t.co/dkc7GDke9Q

[Vishal @learnedquietly]

Any parents here?

This is how our everyday interactions go:

A small thing turns into a tantrum.
You say no. They push back.
You react. It escalates.
Later you think, “I could’ve handled that better.”
But next time… it plays out the same way again.

Sounds familiar?

Every parent goes through this. I did — and I still do.

Most of the time, we try to change our kids.
But maybe it’s the other way around.
Maybe we, as parents, need to change our approach.

That’s what I built — SmallShifts.

You log moments, and over time it shows patterns and suggests small shifts in how to handle them better next time.

https://t.co/dkc7GDke9Q

[Vishal @learnedquietly]

i get the intent behind this

but calling these the “worst things”
just makes parenting feel like one wrong move ruins everything

most of it isn’t that absolute

it’s the overall environment
how we show up again and again
that actually stays with them

[Barbell Financial 💪🏻💰 @BarbellFi]

The worst things for your kids:

Gentle parenting
Daycare before 3
Not breast feeding
Not reading to them
Screen time before 2
Letting them cry it out
Over-scheduling their days
Being on your phone around them
Yelling at each other in front of them

I still see this everywhere 😔

[Vishal @learnedquietly]

yeah

and the strange part is
you don’t even notice it’s ending

it just slowly becomes
something you look back on...

[CooperBaggs 💰🍞 @edgaralandough]

you only get:
1 summer with them as a baby
3 as a toddler
5 as a child
3 as a preteen
6 as a teen
so let them run barefoot,
turn on the sprinkler,
cut up the watermelon,
chase fireflies under fading golden skies
because the summers of sticky hands, sun tired kids, and slow evenings filled with the sounds of childhood
do not last forever 🤍

[Vishal @learnedquietly]

You think your child is being difficult...

Most of the time
they’re just overwhelmed....

We expect them to stay calm
while we are already irritated..

Micro-shift..

regulate yourself first