Anjney Midha
@AnjneyMidha
Followers
32K
Following
11K
Media
525
Statuses
4K
amp up the bitter lessons
San Francisco, CA
Joined May 2011
why I obsess abt compute scarcity my largest angel/early investments: 2021: Anthropic 2022: ElevenLabs 2023: Sesame, Mistral, Luma 2024: Black Forest Labs, LMArena 2025: OpenRouter, Periodic Labs now add up their annual compute needs for next 24 months $1K to the best guess
302
73
2K
first release i am deeply involved in, really happy with what vibe became after it started as a weekend side project i did for fun. and now working on making devstral the best model combined with it.
Introducing the Devstral 2 coding model family. Two sizes, both open source. Also, meet Mistral Vibe, a native CLI, enabling end-to-end automation. 🧵
13
10
214
Over the last year, we've 10X'd+ supply on our marketplace, powering tens of millions of dollars in bookings. Today, we go a step further – enabling Wander to expand beyond the top 1% of vacation homes and service the entire industry (7M+ units) with the launch of Wander Sites.
6
35
86
Huge kudos for the transparency here. Beating Kimi k2 as well is stellar
Introducing the Devstral 2 coding model family. Two sizes, both open source. Also, meet Mistral Vibe, a native CLI, enabling end-to-end automation. 🧵
0
12
247
SOTA open-source vibe coding from your home.
Introducing the Devstral 2 coding model family. Two sizes, both open source. Also, meet Mistral Vibe, a native CLI, enabling end-to-end automation. 🧵
25
51
839
when a team puts an intense amount of love into something they ship, you can feel it in every part of the product
Introducing the Devstral 2 coding model family. Two sizes, both open source. Also, meet Mistral Vibe, a native CLI, enabling end-to-end automation. 🧵
2
7
144
I read a blog post from a chap hating on another startup The hated co just ignored it This made the hater look weak and zero sum, and the hated look stronger Don’t be weak and zero sum. That is for losers and mercenaries Be hated for how strong and mission obsessed you are
3
0
32
earlier this year, i asked jensen how he designed nvidia's org chart 'the org looks like a computing stack' most founders treat org charts as HR paperwork. jensen treats them as architecture. there is a reason nvidia has one of the highest market caps per employee
15
47
528
Ministral is tiny but mighty
Addendum: I ran the Ministral models. The Ministral models are the best scoring open source models released by a major lab to date, and are small enough to be suitable for home inference. Thanks, Mistral.
1
3
22
The righteous man is beset on all sides by the Inequities of the selfish and the tyranny of evil men. Blessed is he who, in The name of charity and good will, shepherds the weak through the valley of Darkness, for he is truly his brother's keeper and the finder of lost Children.
2
0
22
@AnjneyMidha @MaikaThoughts @xanderatallah @cclark One finding: we observe a Cinderella "Glass Slipper" effect for new models. Early users a new LLM either churn quickly or become part of a foundational cohort, with much higher retention than others. They are early adopters who can "lead" the rest of the market (more details 👇)
1
3
32
State of AI: An Empirical 100 Trillion Token Study with @OpenRouterAI, led by @MaikaThoughts @xanderatallah @cclark There are some surprises in here...
7
38
293
We collaborated with @a16z to publish the **State of AI** - an empirical report on how LLMs have been used on OpenRouter. After analyzing more than 100 trillion tokens across hundreds of models and 3+ million users (excluding 3rd party) from the last year, we have a lot of
38
141
772
in early '23, claude had just finished training, and we had to find the first customers for anthropic, fast if you look up the first ever Claude launch blog post, it essentially came down to startup founders like @vivianmshen and @akothari evangelizing it, not 'enterprises'
A startup told me that one of their investors didn't like it that they were selling to newly founded startups, and wanted them to sell to bigger companies, who have more money. If investors tell you this, write them off as idiots. Selling to startups is the best thing you can do.
12
12
186
if you’re interested in autonomous science at frontier scale, come find the @periodiclabs team at neurips! look for: @vwxyzjn to discuss training big MoEs @xiangfu_ml for atomic GNNs @mzhangio for AI scientists @VincentMoens for RL systems me for midtraining sample efficiency
9
12
273
I want to especially thank @MistralAI for releasing the base models for Mistral 3. Fewer companies are sharing base models and this opens many use cases from custom instruct to non-instruct cases
6
16
196
one of the most incredible things about working in tech is the privilege to wake up every day and simply choose not to be zero sum some organizations like @ycombinator have this ethos so deeply embedded in their DNA you feel it in every person you meet from there
5
5
83