swyx #DevWritersRetreat
@swyx
Followers
133K
Following
131K
Media
8K
Statuses
74K
achieve ambition with intentionality, intensity, & integrity - @dxtipshq - @sveltesociety - @aidotengineer - @latentspacepod - @cognition + @smol_ai
san francisco / singapore
Joined April 2009
💁♂️ How to Play Long Term Games: Systems > Goals Discipline > Motivation Trust > Distrust Principles > Tactics Writing > Reading Vulnerability > Confidence North Stars > Low Hanging Fruit Trends > News Habits > Sprints Questions > Answers Problems > Solutions People > Projects
81
805
5K
i was skeptical when @simonw said that "Claude Skills are awesome, maybe a bigger deal than MCP" buuut early indications are this is correct. this is the fastest talk ever to pass 100k views here at AIE. its like those 0 - 100m ARR charts but for attention. @MaheshMurag and
Next up! Claude skills are under-appreciated! @barry_zyj and @MaheshMurag from @AnthropicAI are back at AIE stage this year, to tell us about telling Claude about new skills via simple markdown and code files!
2
0
5
🆕 We're back with a trio of RL talks! @willhang_ and @cathyzbn on OpenAI RFT: https://t.co/HsHlsx4kjz
@willccbb on RL Envs at Scale: https://t.co/3fyK2nQqp5
@rhythmrg and @lindensli giving @AppliedCompute's first ever public talk https://t.co/LFva4Ddruj Our RL track at AIE
0
1
6
Reminder that my course (1000 slides with screencasts) and my "little book" are available under Creative Commons. https://t.co/6OVyjPdwrC
"François Fleuret has 'The Little Book of Deep Learning' and then he also has a full course that he's published. I went through the entire course... I would recommend anybody does this. It goes through the history of deep learning, the topology, the linear algebra, the calculus,
5
18
217
Devstral 2 beats or ties Deepseek v3.2 71% of the time by third party preference and is smaller/faster/cheaper (esp Small 2) !!! i think there's a good chance @theo spoke too soon -Mistral is just starting out on a very epic comeback. Nextgen will get >10x compute and i'd bet
Introducing the Devstral 2 coding model family. Two sizes, both open source. Also, meet Mistral Vibe, a native CLI, enabling end-to-end automation. 🧵
10
18
232
A big ‘complaint’ about NeurIPS circulating on Twitter is that the conference is becoming diluted by non-researchers. In 2023, I attended my first @NeurIPSConf with zero research experience. I was a new grad and the first hire at Exa AI, running around attempting to convince
10
8
150
AIEWF bet big and hard on MCP this year and we were so glad to be joined by @chu_onthis, John Welsh and many other core contributors and big builders! https://t.co/jI5oEleYQI
youtube.com
The winner of the 2023-2025 "agent open standard" wars has conquered the ecosystem — every client from Anysphere to Zed, and every lab from OpenAI to Google ...
1
0
4
proud to have had @aidotengineer help a small part of MCP’s journey
Anthropic is donating the Model Context Protocol to the Agentic AI Foundation, a directed fund under the Linux Foundation. In one year, MCP has become a foundational protocol for agentic AI. Joining AAIF ensures MCP remains open and community-driven.
10
4
109
Hey can we get a 2025 update on this paper pls?
detecting AI-generated text researchers studied the impact of ChatGPT on AI conference peer reviews, confirming what we all knew paper: https://t.co/Dlj7AEJI0h ↓ read more
4
1
24
second confirmation
The Wall Street Journal on the Code Red at OpenAI: When Sam Altman declared a “code red” last week to fend off a rising threat from Google, his first move was not launching a new model or product. Instead he ordered the world’s most valuable startup to pause side projects like
0
0
3
👇 a curated superthread of resources to get the most out of coding agents, advanced context engineering, research/plan/implement, and more
9
20
256
news.smol.ai
As **ICCV 2025** begins, **DeepSeek** releases a novel **DeepSeek-OCR** 3B MoE vision-language model that compresses long text as visual context with high accuracy and efficiency, challenging...
Introducing Nested Learning: A new ML paradigm for continual learning that views models as nested optimization problems to enhance long context processing. Our proof-of-concept model, Hope, shows improved performance in language modeling. Learn more: https://t.co/fpdDlYaleL
0
0
6
agree continual learning very top of mind, but people only seem to have questions and no strong answers other notable neurips takes (see also my subagent RLaaS comments earlier this week) - researchers* surprisingly dont care about: - Deepseek OCR - Muon/MuonClip -
My takeaways from Neurips 1. Continual learning. To support this next frontier, we’re going to need new architectures, new reward functions, new data sources, and new revenue models. 2. Neolabs. Frontier research for risky bets is being shared across multiple companies now 3.
25
15
313
🚢 Capstone of @NeurIPSConf was #ModelShip25! - model lab tea - great food and drinks - beautiful views of the SD Bay - DJ that made AI nerds actually DANCE Incredibly well organized by @caithrin and @ellehcimnehs, and powered by @outshiftbycisco, @LambdaAPI, and @DecibelVC.
10
3
54
i used to have an open office policy - the deal was that you can book me for a codl chat but you have to opt in to me recording and posting them up these things can have a 3 year impact cycle :)
6
2
37
We’re very excited to show reactions to the models for the first time and dig even deeper into the vision. If you’re a top researcher in world modeling, policy learning or have exceptional achievements in engineering or research, please reach out to us. (Disclaimer: we cannot
From a decade building Medal into a 12M-user game clipping platform with billions of highlight moments, Pim turned down a reported $500M offer from OpenAI and raised a $134M seed from Khosla to spin out General Intuition, betting that world models trained on peak human gameplay
5
8
59
From a decade building Medal into a 12M-user game clipping platform with billions of highlight moments, Pim turned down a reported $500M offer from OpenAI and raised a $134M seed from Khosla to spin out General Intuition, betting that world models trained on peak human gameplay
2
5
42