Mehul Mohan's banner

Mehul Mohan

@mehulmpt • 32,362 subscribers

playing with AI research/RL environments at @TachyonAILab. most tweets are notes to self, some for fun. prev founder @fermionapp (acq)

Shorts

coding is solved. my codex desktop app just did a dance for me to celebrate that.

coding is solved. my codex desktop app just did a dance for me to celebrate that.

123,973 görüntüleme

distributed systems are hard

distributed systems are hard

311,779 görüntüleme

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

I asked Astra (GPT-6) and Fable 5.1 both to build GTA VI like game researching about it, and the results are below. Same prompt. Fable took ~2 hours; Astra took ~90 minutes. Watch the full gameplay side by side below, with audio switching every 10 seconds

I asked Astra (GPT-6) and Fable 5.1 both to build GTA VI like game researching about it, and the results are below. Same prompt. Fable took ~2 hours; Astra took ~90 minutes. Watch the full gameplay side by side below, with audio switching every 10 seconds

320,142 görüntüleme • 12 gün önce

Astra (GPT-6) is a full-stack developer that can finish coding work, then verify output, then mark linear tickets as done using your own hardware It's so over

Astra (GPT-6) is a full-stack developer that can finish coding work, then verify output, then mark linear tickets as done using your own hardware It's so over

47,128 görüntüleme • 12 gün önce

Generally, free AI comes with weird limits. However, with this stealth AI model, Ox Alpha + OpenCode Zen gateway, I can personally verify that it is: ✅ Almost unlimited running our internal cybersec bench (I burned 4 billion tokens in 4 hours) ✅ Decent speed ✅ Intelligent model (solves a lot of our internal cybersec tasks) ✅ No cybersecurity guardrails My wild guess is this could be Gemini 3.5 Pro. Full review below

Generally, free AI comes with weird limits. However, with this stealth AI model, Ox Alpha + OpenCode Zen gateway, I can personally verify that it is: ✅ Almost unlimited running our internal cybersec bench (I burned 4 billion tokens in 4 hours) ✅ Decent speed ✅ Intelligent model (solves a lot of our internal cybersec tasks) ✅ No cybersecurity guardrails My wild guess is this could be Gemini 3.5 Pro. Full review below

39,953 görüntüleme • 26 gün önce

Cursor dropped a ₹649/month (~$6.5) plan that includes Grok 4.5 and Composer 2.5. ✅ No 5-hour/weekly limits. ✅ ~25M tokens/month ✅ Grok 4.5 medium (grok 4.5 is GOOD daily driver) ✅ Cursor cloud agent, CLI, or IDE - pick whatever you want. Full review + task:

Cursor dropped a ₹649/month (~$6.5) plan that includes Grok 4.5 and Composer 2.5. ✅ No 5-hour/weekly limits. ✅ ~25M tokens/month ✅ Grok 4.5 medium (grok 4.5 is GOOD daily driver) ✅ Cursor cloud agent, CLI, or IDE - pick whatever you want. Full review + task:

39,433 görüntüleme • 1 ay önce

For the first time in 16 years, a new HTTP method (like POST) is added: Method QUERY. This is cacheable by CDNs, like GET requests, but allows a body in the request. Talking about this in detail today, and a hot take on this at the end.

For the first time in 16 years, a new HTTP method (like POST) is added: Method QUERY. This is cacheable by CDNs, like GET requests, but allows a body in the request. Talking about this in detail today, and a hot take on this at the end.

35,968 görüntüleme • 2 ay önce

I tried Kimi K3 Max on the $99/month 1 million context window plan, and shipped a micro SaaS. Here are my learnings with a full breakdown

I tried Kimi K3 Max on the $99/month 1 million context window plan, and shipped a micro SaaS. Here are my learnings with a full breakdown

22,522 görüntüleme • 2 ay önce

I believe 10 years down the line it would not be uncommon to buy AI intelligence in a new home, for an extra $400-$500k. A centralised intelligence box/rack that every AI harness, including hardware/robots would use. Talking about that future a bit in a video today.

I believe 10 years down the line it would not be uncommon to buy AI intelligence in a new home, for an extra $400-$500k. A centralised intelligence box/rack that every AI harness, including hardware/robots would use. Talking about that future a bit in a video today.

26,112 görüntüleme • 2 ay önce

Introducing Fermion - the most powerful way to run your online education business. As a past ed-tech startup, the biggest mistake we made was spending too much time building technology, rather than focusing on our unique strengths and user outcomes. Fermion changes that.

Introducing Fermion - the most powerful way to run your online education business. As a past ed-tech startup, the biggest mistake we made was spending too much time building technology, rather than focusing on our unique strengths and user outcomes. Fermion changes that.

35,929 görüntüleme • 10 ay önce

Your website on replit/lovable, etc. is guaranteed to be 4-10x slower than cloud platforms like cloudflare/vercel that deliver from their edge locations. To clarify, it sounds weird to me how these vibe AI platforms just position themselves as a "Vercel" / "Cloudflare" alternative

Your website on replit/lovable, etc. is guaranteed to be 4-10x slower than cloud platforms like cloudflare/vercel that deliver from their edge locations. To clarify, it sounds weird to me how these vibe AI platforms just position themselves as a "Vercel" / "Cloudflare" alternative

26,766 görüntüleme • 11 ay önce

Daha fazla içerik yok.