LinkedIn Content Strategy & Writing Style
Infra Investing at Moxxie Ventures | Author of 13 AI books | Nvidia alum | Recovering Founder
0 people tracking this creator on ViralBrain
Prateek Joshi positions himself as a high-signal bridge between deep academic research and venture-scale infrastructure investing. Leveraging his background as a 13-time author and Nvidia alum, his content strategy centers on distilling low-level technical bottlenecks-such as kernel engineering and KV cache optimization-into actionable insights for the startup ecosystem. He is notable for his ability to translate complex MLSys research papers into a coherent thesis on where the next "bottleneck of value" will accumulate. By maintaining a unique intersection of hardware-aware systems engineering and venture capital, Prateek provides a rare "field report" perspective that moves beyond high-level AI hype to focus on the gritty, verifiable physics of scaling intelligence.
0
0
6
—
5.0
9
15
Claude Code source code got leaked. And someone looked up the one thing that truly matters: Spinner verbs. Turns out there are 187.
Researchers at Google just published a fascinating finding: Simply repeating your prompt twice can dramatically improve LLM performance. In some cases taking accuracy from 21% to 97% on specific searc…
A tale of two cities. There are two opposing views here: 1. Pure vibe coding will lead to a lot of technical debt, which is bad. So you need to learn how things work. 2. AI models are getting so go…
Training LLMs has historically required tightly controlled GPU clusters, but that assumption may be starting to crack. Templar just completed what they claim is the largest decentralized LLM pre-tra…
For centuries, mathematics has relied on a simple social contract: a mathematician writes a proof, and other mathematicians check it. If the reasoning holds up, the result becomes part of the canon.…
"Pure software is rapidly becoming un-investable" But my pushback to this is that generating code was never the true bottleneck. When the cost of building something collapses, demand expands. Cloud…
5.0 posts/week
Posts / Week
1.6 days
Days Between Posts
15
Total Posts Analyzed
HIGH
Posting Frequency
5.9%
Avg Engagement Rate
STABLE
Performance Trend
850
Avg Length (Words)
HIGH
Depth Level
ADVANCED
Expertise Level
0.9/10
Uniqueness Score
NO
Question Usage
0.2%
Response Rate
Writing style breakdown
<start of post>
The economics of inference are shifting from "cost per million tokens" to "cost per successful task completion."
As models move from simple chat to long-running agents, the bottleneck is no longer just raw throughput. It is the ability to maintain state, manage memory, and coordinate tool calls over thousands of steps without drifting.
This is why we are seeing a surge in research around KV cache compression and tiered memory architectures.
Memory-as-a-service: decoupling KV cache from compute nodes
Speculative execution for agentic tool calls
Dynamic context window resizing based on task complexity
Hardware-aware scheduling for heterogeneous agent swarms
So what does this mean for the stack?
We are moving towards an "Agentic Operating System" where the LLM is just the CPU, and the surrounding infra handles the memory, scheduling, and I/O.
The signal for founders: don't just build a faster engine. Build the memory and orchestration layer that allows that engine to run for days, not seconds.
cc: Moxxie Ventures, Alex Roetter
<end of post>
Sign in to unlock the full writing analysis
Free tools to help you write, score, and benchmark in the same style.
Other creators worth studying alongside Prateek Joshi.
Jason Fairchild
Co-Founder and CEO at tvScientific
103 Viral Score
Fabrice Courdesses
Co-Founder & CEO of VideoRunRun | Data-Driven Creative Intelligence Platform for Video Ads Performance
10 Viral Score
Nicolas JAIMES
Founder of Open Garden
117 Viral Score
Richard Mulholland
Founder: Missing Link and Too Many Robots. We help companies be a threat to the future (before the future threatens them)
31 Viral Score
Miguel Otero Pedrido
ML / AI Engineer | Founder @ The Neural Maze - Just a guy who builds AI Systems that actually work
29 Viral Score
Daniel Moka
I help you craft better software
95 Viral Score
ViralBrain plans, writes, and schedules your LinkedIn content — using official LinkedIn APIs so your account stays safe.
Write like Prateek Joshi.