Every time you send a prompt to an AI app, something behind the scenes has to decide which model inference server actually answers it. Most people never think about this. But that decision — the “router” — is quietly turning into one of the most important pieces of the AI stack. This is a walkthrough of what routers actually do today, what’s still shallow about them, and where the smart money says they’re headed next.
🚴🎧 over 2x rides. Getting rack scale system built in this short of time is no joke. Beats all the BS peddled where silicon is only for grey haired. https://t.co/Jv8KOWeyrx
— Abhay 🇸🇬🇮🇳 (@Abhay08)
Jul 2, 2026
TPU Deep dive
https://t.co/lKvhLNDZ4A
— Abhay 🇸🇬🇮🇳 (@Abhay08)
Jun 29, 2026
#3 Token prices are heading to ~1/10th of today’s.
Compute scarcity props prices up while consumers (unprofitable) eat half the supply.
As efficiency catches up: prices collapse, usage explodes – Jevons paradox. The durable moat becomes memory + context aka The Harness, not the— Abhay 🇸🇬🇮🇳 (@Abhay08)
Jun 26, 2026
#2 Consumers tolerate false positives. Enterprises tolerate zero.
That one gap explains the whole market.
Frontier models chase consumer breadth (brand + post-training data). Real agentic enterprise value needs depth, context, memory — that’s where the moat forms.— Abhay 🇸🇬🇮🇳 (@Abhay08)
Jun 26, 2026
https://t.co/qMSxphnGU9 The breadth of topics and how masterfully Nikesh navigates is just fabulous.
#1 Physical AI is a depth problem, not breadth.
No consumer use case = no crowdsourced training data. The model that flies planes ≠drives cars ≠runs factories.
Waymo proves— Abhay 🇸🇬🇮🇳 (@Abhay08)
Jun 26, 2026
Nails the FDE craze the industry is going through now https://t.co/qKHl7YoqNh
— Abhay 🇸🇬🇮🇳 (@Abhay08)
Jun 16, 2026
Unsecured leveraged lending does not exist for retail. Most are either securities-backed or mortgage-backed, but 7x to 10x leverage is unheard of. On top, there is debt service ratios levied by banks. @deepakshenoy views? https://t.co/pzUcMAujuF
— Abhay 🇸🇬🇮🇳 (@Abhay08)
Jun 12, 2026
@Ankit_Quant Most so-called Fin-Influencers are doing these clickbait posts saying 20% IRR return on revised FCNR deposit rates across 3-5 Yrs. Pure BS for 80-90% of NRI’s. No FI provides that kind of leverage to retail folks. Shows how shallow the influencers are
— Abhay 🇸🇬🇮🇳 (@Abhay08)
Jun 12, 2026
RT @paulg: If big companies can’t make a net return on their LLM token costs, that doesn’t mean it’s impossible to. In fact this is exactly…
— Abhay 🇸🇬🇮🇳 (@Abhay08)
Jun 7, 2026