Dharma

stocks, cricket and bit of humor

[Dharma @dharmartho]

@rajmafi @sumanthraman Blame the govt for not listening to its core supporters 3 months back and then crumbling to the street.

[Abhishek Kothari 🇮🇳 @kothariabhishek]

#Whatsapp forward

Reliance: No return in 4 yrs.

HDFC Bank: No return in 5 yrs.

TCS: No return in 7.75 yrs.

HUL: No return in 6.5 yrs.

Infosys: No return in 5.5 yrs.

Maruti Suzuki: No return in 2 yrs.

Adani Ent: No return in 3.75 yrs.

Axis Bank: No return in 1.5 yrs.

M&M: No return in 1.75 yrs.

Kotak Bank: No return in 5.5 yrs.

ITC: No return in 9 yrs.

NTPC: No return in 2 yrs.

ONGC: No return in 12 yrs.

Ultratech: No return in 2 yrs.

HCL Tech: No return in 5.75 yrs.

Coal India: No return in 2.5 yrs.

Bajaj Auto: No return in 1.75 yrs.

HAL: No return in 2 yrs.

Bajaj Finsv: No return in 4.75 yrs.

Dmart: No return in 4.75 yrs.

Nestle: No return in 2.5 yrs.

Powergrid: No return in 2.5 yrs.

Asian Paints: No return in 5.5 yrs.

Adani Green: No return in 4.5 yrs.

Eternal: No return in 1.75 yrs.

Hind Zinc: No return in 2 yrs.

Wipro: No return in 5.5 yrs.

Indian Oil: No return in 9.25 yrs.

Adani Energy: No return in 5 yrs.

SBI Life: No return in 1.75 yrs.

Varun Bev: No return in 2.5 yrs.

Indigo: No return in 2 yrs.

Jio Fin: No return in 2.75 yrs.

Trent: No return in 2.25 yrs.

Tata Motors: No return in 2.75 yrs.

Tech M: No return in 4.75 yrs.

Cipla: No return in 2.5 yrs.

Tata Cons: No return in 2.5 yrs.

Dr Reddy: No return in 2.5 yrs.

Max Health: No return in 2.5 yrs

[Morty @0xMortyx]

Andrej Karpathy just broke the entire premise of modern AI:

"Agents aren't magic. They're distillation at scale."

99.99% of your LLM's capacity is wasted on garbage data it never needed.

Small model + right tools + closed loop = terrifying capability.

In a 16-minute conversation, Karpathy reveals the full reasoning stack.

Worth more than any $500 AI course you've seen this year.

[h100envy @h100envy]

Google engineer explained how to fine-tune a tiny LLM from 46% to 90% accuracy on your phone in 21 minutes - better than $1500 on-device AI bootcamps.

pick Gemma 270M -> generate synthetic task data -> fine-tune with LoRA -> quantize to int4 -> deploy to Pixel and hit 2000 tokens per second.

That loop is how a 270M model beats a 70B one on your task, running fully offline in your pocket.

Gemma 270M + synthetic data + LoRA + int4 quantization + on-device runtime - that's the stack.

Watch and save it, then fine-tune your own tiny agent tonight.

[Carnage @0xCarnagee]

Prime Intellect engineer:

"everyone's bragging about a million-token context. here's what they don't tell you.

at 256k tokens GPT-5.5 scores 80% on retrieval. push it to a million and it drops to 36%. the model accepts the context, it just can't reason across it. people call it context rot."

in a 20-minute talk he explains why bigger context windows won't save your agents.

continual learning + training on your own traces + real environments - that's the fix.

Watch the talk, then save!

[Puneet Patwari @system_monarch]

Suppose you have interviews scheduled for these 70LPA-1Cr+ CTC roles:

Google L5 /
Meta E5 /
Uber Staff /
Amazon L6 /
Salesforce SMTS /

Go through these 27 core system design concepts and problems.

How many can you reason through in a 1-hour round if the interviewer injects one of them into your design or asks as a follow-up? How much clarity do you have?

Beginner:
- The thundering herd problem
- Cache stampede
- N+1 query problem
- Hot partition / hot key
- Single point of failure
- Retry storm
- Backpressure
- Duplicate requests / idempotency gap
- Stale cache / read-after-write inconsistency

Intermediate:
- Distributed rate limiting
- Leader election
- Distributed locking and lease expiry
- Quorum reads vs quorum writes
- Fan-out on write vs fan-out on read
- Out-of-order event processing
- Dead letter queues and poison messages
- Zero-downtime schema migration
- Circuit breaker and cascading failure control

Advanced:
- Split a monolith safely
- Multi-region failover
- Active-active conflict resolution
- Change Data Capture vs dual writes
- Search index freshness vs ranking quality
- Rebalancing shards under skewed traffic
- Noisy neighbor problem in multi-tenant systems
- Watermarks / late-arriving events in stream processing
- Exactly-once processing vs practical deduplication

Most candidates prepare for:
- “Design Uber.”
- “Design Twitter.”
- “Design Dropbox.”

These generic prompts are good for beginners, but at senior levels, depth matters. Strong candidates prepare for the even minute problems that can cause disasters at scale.

[Jack @4nsdp]

Software engineers: Do you know why JWTs start with ey? 🤔

[Puneet Patwari @system_monarch]

I've been a backend Engineer for 12+ years. Today, I'm a Principal Engineer at Atlassian.

I've designed systems that handle millions of requests. Sat on both sides of system design interviews.

Reviewed more architecture docs than I can count.

Starting today, I'm breaking down the fundamentals of scaling for the next 25 days.

If you're learning system design bookmark this thread, you're going to get a lot of learning from this.

[Dharma @dharmartho]

@vasloga thanks bro

[Vinodsrinivasan @vinodsrinivasan]

TOI's "3-year backstory behind PM's appeal" is the cleanest balance of payments picture I've seen in mainstream Indian media.

Rupee weakness is not a 2026 story. It is a structural problem dressed up as a crisis.

8 charts. Let me walk you through what's actually happening. 🧵

[Dharma @dharmartho]

@Iam_No_One____ @sumanthraman @iamshalabh21 AI needs so much to actually be efficient as perceived, you running on a bunch of files and assuming it's all perfect shows ur bias.
Try scraping through datasets which are generated 1 tb a day with explainability on why it did whatever it did u will know how difficult it is

[Dharma @dharmartho]

@Iam_No_One____ @sumanthraman @iamshalabh21 possibley imagine?

Dude ur team built it for u, I have built multi modal agents which are capable of reasoning and then decision making based on the ifnerence or to HITL.

My point is with AI, the largest missing piece is data which will keep evolving and AI cannot solve that.

[Dharma @dharmartho]

@Iam_No_One____ @sumanthraman @iamshalabh21 U have any idea on how AI works before making such statements?

[Out Of Context Cricket @GemsOfCricket]

RCB fans watching Venkatesh Iyer batting

[Sofia khatun @sofia3973]

🚨 BREAKING: These 15 careers will quietly dominate the next 10 years.

Most people won’t notice…
until the money, leverage, and opportunities are gone.

Use Claude to learn these early. 👇

[Puneet Patwari @system_monarch]

I'm a Principal with 12 years of experience. If I was coaching you to crack system design rounds for Sr to Staff+ AI/ML roles at companies like Meta, Google, Salesforce, Amazon, etc.

I would 100% ask you to work on these fundamentals before we would start talking about interviews.

Because AI system design is still system design.

The only difference is that now your bottlenecks are not just databases, caches, and queues.

They are tokens, context windows, retrieval quality, inference cost, hallucinations, model latency, evals, and user trust.

Here are the fundamentals I would start with:

➤ LLM Basics
↬ Tokens
↬ Context Window
↬ Prompt Design
↬ System Prompts
↬ Temperature
↬ Top-p Sampling
↬ Structured Outputs
↬ JSON Mode
↬ Function Calling
↬ Tool Calling
↬ Agents
↬ Memory
↬ Guardrails
↬ Hallucinations
↬ Model Latency
↬ Model Routing
↬ Small vs Large Models
↬ Fine-tuning vs Prompting
↬ Open-source vs Closed Models

➤ RAG & Retrieval
↬ Embeddings
↬ Vector Search
↬ Vector Databases
↬ Chunking
↬ Chunk Overlap
↬ Metadata Filtering
↬ Hybrid Search
↬ Keyword Search
↬ Semantic Search
↬ Reranking
↬ Retrieval Recall
↬ Retrieval Precision
↬ Query Rewriting
↬ Document Freshness
↬ Permission-aware Retrieval
↬ Citation Grounding
↬ Evidence Selection
↬ Context Packing
↬ Missing Information Detection

➤ AI System Architecture
↬ API Gateway
↬ Request Routing
↬ Model Gateway
↬ Prompt Service
↬ Inference Service
↬ Retrieval Service
↬ Ranking Service
↬ Feature Store
↬ Offline Pipelines
↬ Online Serving
↬ Async Processing
↬ Queueing
↬ Streaming Responses
↬ Rate Limiting
↬ Fan-out/Fan-in
↬ Batch Inference
↬ Real-time Inference
↬ Human-in-the-loop Systems
↬ Fallback Workflows

➤ Cost & Performance
↬ Token Budgeting
↬ Prompt Compression
↬ Prompt Caching
↬ Semantic Caching
↬ Response Caching
↬ Batch Requests
↬ Model Quantization
↬ Distillation
↬ Latency Budgets
↬ Cold Starts
↬ GPU Utilization
↬ Throughput
↬ Cost per Query
↬ Cost per User
↬ Model Selection
↬ Inference Scaling
↬ Backpressure
↬ Load Shedding

➤ Evaluation & Quality
↬ Offline Evals
↬ Online Evals
↬ Golden Dataset
↬ Human Review
↬ LLM-as-Judge
↬ A/B Testing
↬ Regression Testing
↬ Answer Relevance
↬ Factual Accuracy
↬ Faithfulness
↬ Groundedness
↬ Toxicity Checks
↬ Safety Checks
↬ Drift Detection
↬ Feedback Loops
↬ Confidence Scoring
↬ Escalation Criteria
↬ Quality Monitoring

➤ Reliability & Security
↬ Timeouts
↬ Retries
↬ Circuit Breakers
↬ Failover
↬ Model Fallbacks
↬ Graceful Degradation
↬ Observability
↬ Tracing
↬ Prompt Logs
↬ Token Metrics
↬ Error Budgets
↬ PII Redaction
↬ Data Privacy
↬ Access Control
↬ Prompt Injection
↬ Jailbreak Defense
↬ Audit Logs
↬ Compliance

[Akshay Shinde @ConsciousRide]

90% of LangGraph & Agent Framework interviews in 2026 are just these 10 concepts repeated:

[Ryan @WritesToProfit]

If I were in my 30s or 40s & wanted to retire in the next 10 years using AI, here’s exactly what I’d do:

1. Set up an LLC immediately. Not next month. Not after you "feel ready." This week.

[Avinash Singh @AvinashSingh_20]

Bangalore Jobs Seekers Kit - https://t.co/CDMuTHARwR

Gurgaon Job Seekers Kit - https://t.co/leAPq2sfMk

Hyderabad Job Seekers Kit - https://t.co/bAeyPk2vKk

Mumbai Job Seekers Kit - https://t.co/yixwNn2rJg

Bookmark it!