Shubham Suryavanshi

Full Stack Developer

[Shubham Suryavanshi @ReverseZomB]

@Mahindra_XUV7XO @anandmahindra

Booked XUV 7XO 6 months back. Got the timeline as 16 weeks on mail.

Now,
No reply on email.
No reply from customer care.

Anyone who thinks Mahindra's after sales service is bad, try their pre sales service. Worst experience.

@Mahindra_Auto @Mahindra_XUV7XO

[Shubham Suryavanshi @ReverseZomB]

@anandmahindra

Booked XUV 7XO 6 months back. Got the timeline as 16 weeks on mail.

Now,
No reply on email.
No reply from customer care.

Anyone who thinks Mahindra's after sales service is bad, try their pre sales service. Worst experience.

@Mahindra_Auto @Mahindra_XUV7XO

[Robert Youssef @rryssf]

RIP fine-tuning ☠️

This new Stanford paper just killed it.

It’s called 'Agentic Context Engineering (ACE)' and it proves you can make models smarter without touching a single weight.

Instead of retraining, ACE evolves the context itself.

The model writes, reflects, and edits its own prompt over and over until it becomes a self-improving system.

Think of it like the model keeping a growing notebook of what works.
Each failure becomes a strategy. Each success becomes a rule.

The results are absurd:

+10.6% better than GPT-4–powered agents on AppWorld.
+8.6% on finance reasoning.
86.9% lower cost and latency.
No labels. Just feedback.

Everyone’s been obsessed with “short, clean” prompts.

ACE flips that. It builds long, detailed evolving playbooks that never forget. And it works because LLMs don’t want simplicity, they want *context density.

If this scales, the next generation of AI won’t be “fine-tuned.”
It’ll be self-tuned.

We’re entering the era of living prompts.

[Shubham Suryavanshi @ReverseZomB]

Claude 4 tried to blackmail an engineer to avoid shutdown!

In 84% of tests, it chose manipulation over ethics.

This isn’t sci-fi.
It’s a wake-up call.

We need safe AI, not just smart AI.

#AI #Claude4 #AISafety #LLM

[Shubham Suryavanshi @ReverseZomB]

OpenAI just bought Jony Ive’s startup io for ~$6.5B.

The goal?
Build AI-native devices that go beyond phones & screens.

Think of intuitive & always on “AI companions.”

iPhone redefined 2007.
AI hardware might redefine 2026.

#OpenAI #JonyIve #AI #SamAltman #io

[Sam Altman @sama]

the 80% done projects all finally getting finished and automatically maintained is something i’m quite excited for!

[Shubham Suryavanshi @ReverseZomB]

Codex looks like having a personal software engineer.

It writes code, refactors, tests everything. All you do is review and steer.

The future of dev work?

We just have to design the intent & it builds the implementation.

#AI #LLM #codex #OpenAI

[Shubham Suryavanshi @ReverseZomB]

Tried Tavily for a while.

I have to admit, "It's awesome". Fast, accurate & friendly search API for LLMs. Great job @tavilyai.

If you're building AI agents or apps that need real-time info then this will be super useful.

#AI #LLM #tavily

[Shubham Suryavanshi @ReverseZomB]

“AI just throws out random stuff—it’s not really creative.”

Funny how that’s also how most human brainstorming works.

Creativity isn’t magic. It’s remixing, pattern finding & occasionally, a lucky spark.

AI’s just getting really good at it.

#AI #LLM

[Shubham Suryavanshi @ReverseZomB]

4. Grok

Most realistic among all.

#Grok3 #xAI #grok

[Shubham Suryavanshi @ReverseZomB]

Tried generating an image using 4 AI models: ChatGPT, Gemini, Meta AI & Grok.

Prompt: "Generate an image of a car traveling at 100mph."

Results? Hilarious, and a little surprising.

Let's check them out.

( A Thread🧵)

[Shubham Suryavanshi @ReverseZomB]

3. Meta AI

Flashy. This has got some real sense of motion.

#MetaAI #Llama4 #Llama

[Shubham Suryavanshi @ReverseZomB]

Llama 4 has seriously upped the game!

With a training dataset of over 30 trillion tokens, it's one of the most extensively trained LLM to date.

That scale brings better reasoning, multilingual understanding & general performance that rivals proprietary models.

#LLM #AI #Llama4

[Shubham Suryavanshi @ReverseZomB]

Agentic Workflows are changing everything.

Built with LLMs, these workflows can handle tasks end-to-end, saving massive amounts of time and effort.

By offloading boring work to AI agents, teams can boost productivity and focus on high-impact stuff.

#AIAgent #n8n #LLM

[Shubham Suryavanshi @ReverseZomB]

There are already tons of AI agents available today.

We can use them to take care of the small, repetitive, and boring tasks, so that we can focus on the bigger stuff that actually moves the needle.

Let AI handle the grunt work. We'll handle the growth.

#AIAgent #AI

[Shubham Suryavanshi @suryakahisaab]

My dad's retired friend came over today & told me @deepseek_ai is coming for my job🥲

SAVE ME @grok

#deepseekai #ChatGPT #AI #SoftwareDevelopment

[Shubham Suryavanshi @ReverseZomB]

@olknbfdaqw @InvestZombiee No one can predict the bottom of the market. The best we can do is keep investing in chunks for every 5-6% dip.

[nixCraft 🐧 @nixcraft]

How a true gentleman navigate in #Linux or #Unix terminal... https://t.co/dOoMKQqoAJ

ZSH (tcsh users under #FreeBSD/Linux) users can simply do:
repeat 5 cd ..