Deedy

@deedydas on X

Writes about AI, Engineering and Revenue. 68% of posts open with an observation. A typical post reaches 1,581 likes.

Remix their style

Posts: 38
Likes: 113K
Typical: 1.6K
Best: 19K

What they write about

AI: 87%
Engineering: 29%
Revenue: 21%

How they open

observation: 68%
number: 26%
announcement: 5%

Every post of theirs in the library

Deedy@deedydas

If you haven’t used Grok 3 “Sexy” 18+ mode, drop everything and try it. I can’t explain how unbelievably messed up this is (and I can’t post a video). This may single handedly bring down global birth rates. I can’t believe Grok actually shipped this.

Opens with a number

Deedy@deedydas

BREAKING DeepSeek just let the world know they make $200M/yr at 500%+ profit margin. Revenue (/day): $562k Cost (/day): $87k Revenue (/yr): ~$205M This is all while charging $2.19/M tokens on R1, ~25x less than OpenAI o1. If this was in the US, this would be a >$10B company.

Opens with a number

Deedy@deedydas

India's biggest AI startup, $1B Sarvam, just launched its flagship LLM. It's a 24B Mistral small post trained on Indic data with a mere 23 downloads 2 days after launch. In contrast, 2 Korean college trained an open-source model that did ~200k last month. Embarrassing.

Opens with a number

Deedy@deedydas

Manus, the new AI product that everyone's talking about, is worth the hype. This is the AI agent we were promised. Deep Research+Operator+Computer Use+Lovable+memory. Asked it to "Do a professional analysis of Tesla stock " and it did ~2wks of professional-level work in ~1hr!

Opens with an observation

Deedy@deedydas

PDF parsing is solved (again). Mistral's new OCR API — parses 1000-2000 pages for $1 — achieves state of the art results on tables, multilingual — supports structure: images, bounding boxes, scans, equations 90% of the world's organizational data is in PDFs.

Opens with an observation

Deedy@deedydas

China just dropped a new model. ByteDance Doubao-1.5-pro matches GPT 4o benchmarks at 50x cheaper — $0.022/M cached input tokens, $0.11/M input, $0.275/M output — 5x cheaper than DeepSeek, >200x of o1 — 32k + 256k context — sparse MoE architecture AI truly too cheap to meter.

Opens with an observation

Deedy@deedydas

We just solved text-to-speech AI. This model can simulate perfect emotion, screaming and show genuine alarm. — clearly beats 11 labs and Sesame — it’s only 1.6B params — streams realtime on 1 GPU — made by a 1.5 person team in Korea!! It's called Dia by Nari Labs.

Opens with an observation

Deedy@deedydas

Make no mistake, Google's new Veo 3 video generatiom model is absolutely exceptional. But gymnastics is still pure nightmare fuel, and is the Turing test for video models!

Opens with an observation

Deedy@deedydas

The ex-quants in China dropped the best open-source LLM in the world: DeepSeek V3! — 671B params, 37B active MoE — On par with 3.5 Sonnet and 4o — $0.27/Mtok input, $1.1/Mtok output — 60tok/s — 128k context — Trained with just $5.5M on 14.8T tok! 53-page technical paper is GOLD

Opens with an observation

Deedy@deedydas

Google really cooked with Gemini 2.0 Flash Thinking. It thinks AND it's fast AND it's high quality. Not only is it #1 on LMArena on every category, but it crushes my goto Math riddle in 14s—5x faster than any other model that can solve it! Google is making OpenAI dance.

Opens with an observation

Deedy@deedydas

Wow, Sonnet 3.7 w/ Thinking just solved a problem no other model could solve yet. "Can you write the most intricate cloth simulation in p5js" Grok 3 and o1 pro had no usable results. This is truly the best "vibe coding" model.

Opens with an observation

Deedy@deedydas

This insane result from Claude is the biggest in analytic number theory since bounded prime gaps in 2013. It boosted the proven fraction of Riemann zeta zeros on the critical line by 25.6pts. In the 37yrs prior, mathematicians moved it 0.8pts.

Opens with an observation

Deedy@deedydas

HUGE! OpenAI in talks to buy Codeium for $3B and it's been <4yrs since it's founding. There's clear intent for OpenAI to verticalize and own the app layer, including coding and compete with Cursor head-on.

Opens with an observation

Deedy@deedydas

DeepSeek R1 isn't just "25x cheaper than GPT o1"... It is better than the unreleased OpenAI o3 at the same cost at coding on Codeforces and ARC-AGI!

Opens with a number

Deedy@deedydas

Google's AI models have been racking up the Ws. Best small model: Gemma 27B Best context window: 2M toks Best video understanding: Gemini (all) Best image editing: Gemini 2 Flash Best cost-quality tradeoff: Gemini 2 Flash Best video gen: Veo 2 Best audio gen product: NotebookLM

Opens with an observation

Deedy@deedydas

Meta’s AI org is in disarray. First Soumitra Chintala, the inventor of PyTorch leaves. Now Yann LeCun, their AI head leaves. They have $600B in compute commits until 2028, but I guess it’s up to Alex Wang and Nat Friedman to deliver.

Opens with an observation

Deedy@deedydas

I feel so bad for Meta. They had to push back the Llama 4 release to match DeepSeek R1. Now R2 is announced and may land before Llama 4 which might force them to push it back even more. The US may lose on open source AI to China.

Opens with an observation

Deedy@deedydas

For the first time ever, YouTube's revenue surpassed $50 billion, says Sundar Pichai at the Google earnings. That makes it the #1 highest grossing video app with Netflix, ~$37B, at #2!

Opens with a number

Deedy@deedydas

Excited to announce that we at Menlo Ventures are doubling down on our investment with Anthropic! $61.5B is a small price for 10x'ing every software engineer.

Opens with an announcement

Deedy@deedydas

Anthropic is at $14B run rate revenue, the fastest growing software business of all time. – Claude Code run rate is $2.5B in <1yr – $0 -> $100M -> $1B -> $14B in 3yrs – $100k+ customers 7x'd last year – $1M+ customers 40x'd last year, 500+ – 4% of Github commits by CC Insane.

Opens with a number

Deedy@deedydas

Thinking Machines just dropped the best open weight AI model outside of China! Inkling beats Nemotron 3 Ultra and benchmarks put it between Kimi 2.5 & 2.6. Many were contending to this throne, but Thinky has come out on top. Really solid release, and will pair well with Tinker.

Opens with an observation

Deedy@deedydas

Claude 3.5 is here. Sonnet is the first release and has: — 2x the speed of Opus — 1/5th the cost of Opus — 200k token context window — Better quality than Opus and GPT-4o I don't trust benchmarks so I tried a Physics q that GPT 4-o failed and Sonnet nailed it. Insane launch.

Opens with an observation

Deedy@deedydas

HUGE news about enterprise LLM usage! AI spend in 2024 has gone from $2.3B to $13.8B, a 6x increase. Anthropic with 24% has been catching up to OpenAI's lead at 34% on driven by how its proficiency at code, which makes up >50% the spend! Read the full report from Menlo here:

Opens with an observation

Deedy@deedydas

Summary of all the CRAZY benchmark results from OpenAI's most advanced model, o3! SWE-Bench Verified: 71.7% Codeforces rating: 2727 Competition Math: 96.7% PhD level science (GPQA): 87.7% Frontier Math: 25.2% (previous best was 2%) ARC-AGI: 87.5% TRULY SUPERHUMAN.

Opens with an observation

Deedy@deedydas

BREAKING Anthropic to release Claude Sonnet 3.7 on Feb 26 It’s expected to have step-by-step thinking, never before seen coding capabilities and web search. The best coding model which powers Cursor and Windsurf is about to get a whole lot better.

Opens with an observation

Deedy@deedydas

DeepSeek matches o1 performance at 25-30x cheaper price! Per million input (cached), input and output tokens: o1: $7.5, $15, $60 o1-preview: $1.5, $3, $12 DeepSeek Reasoner: $0.14, $0.55, $2.19 (!)

Opens with a number

Deedy@deedydas

"Thinking Machines' Tinker does a few hundred million of ARR" - Dylan Patel, on his latest podcast TML was valued at $12B and trying to raise at $50B. Tinker is an API that helps you post-train LLMs. This is the highest known revenue for any of the now ~75 neolabs.

Opens with an observation

Deedy@deedydas

We just dropped a 12 page AI report on how ~500 execs at US enterprises use generative AI. I read it all so you don't have to. Top 8 takeaways: Anthropic is the #1 model provider in the enterprise, with 40% of ~$37B spend, with OpenAI dropping to #2. 1/8

Opens with a number

Deedy@deedydas

The future of building software. LLMs are pretty good at generating code, but they're slow. Gemini Diffusion is 10-15x faster than autoregressive models by using diffusion, which used to be for images. This is the 2nd model after Mercury Small to show this. 2/12

Opens with an observation

Deedy@deedydas

Excited to announce that we at Menlo Ventures are leading a $50M round in Goodfire with Anthropic to understand AI models! Having started interpretability teams at OpenAI and DeepMind, we think this is the best research team to solve this. The future of AI is interpretable.

Opens with an announcement

Deedy@deedydas

Meta will delay its biggest AI model launch, Llama 4 Behemoth. —Doesn’t perform well internally —Huge reorg in AI lesdership —11 of 14 researchers on Llama have left —All this after admitting they gamed LMSys Glad Meta can afford to light billions of $$ on fire for open source!

Opens with an observation

Deedy@deedydas

According to the o1 system card, OpenAI’s newest and best model o1 not only underperforms state of the art (Claude Sonnet 3.5) in SWE-Bench for coding but also… Is on par with o1-preview!

Opens with an observation

Deedy@deedydas

Anthropic Sonnet 3.7 just dropped and it's the best coding model in the world right now! It scores 62.3% on 500 problems with pass@1 and 70.3% on a reduced subset with a custom scaffold. And you can try it NOW.

Opens with an observation

Deedy@deedydas

Apple's going to use Anthropic Claude Sonnet to power AI code gen in XCode to develop its own Swift apps! OpenAI is not there yet and using Google is likely a non starter. A huge step for codegen outside the Python / JS ecosystem!

Opens with an observation

Deedy@deedydas

OpenRouter is now serving 1.5 quadrillion tokens/yr! That token run rate is: — 15-30% of Google APIs — 20-40% of OpenAI — >50% of Microsoft Azure Foundry That's 15x larger than when we invested a year ago. Revenue has already doubled since this $1.3B round was done in Feb!

Opens with an observation

Deedy@deedydas

Cybersecurity startup Wiz gets acquired by Google for $32B. Their secret? Elite Israeli cybersecurity seed investors Cyberstarts and their CISO network that help bootstraps their companies. Cyberstarts owns ~4% or ~$1.3B, a >320x return in 4yrs.

Opens with a number

Deedy@deedydas

OpenAI just dropped GPT 4.1 today, "Quasar Alpha" on OpenRouter. — Performance: 55% w/o reasoning on SWE-Bench Verified — Latency: Similar to GPT 4-o — Price: $2/M input tokens, $8/M output — Features: 1M context window, 32k output Intelligence keeps getting cheaper every week.

Opens with an observation

Deedy@deedydas

New 2024 court docs from Facebook’s FTC trial reveal.. WhatsApp isn’t just making $1-2B/yr* from its Business API, but likely ~$2.5-3B/yr now from Click-to-WhatsApp Ads! WhatsApp is now ~3-4% of Meta revenue and is likely #2 highest-earning messaging app after Chinese WeChat!

Opens with a number