Noah Hayes

1.4K posts

Noah Hayes banner
Noah Hayes

Noah Hayes

@ninjoah

Helping businesses help businesses help customers, with help from AI. @vidaisai

San Francisco, CA Katılım Eylül 2008
288 Takip Edilen437 Takipçiler
Nate Berkopec
Nate Berkopec@nateberkopec·
Study finds that skills written by the LLM itself do not increase task completion rate. Definitely matches my own experience. LLMs are noisy amplifiers: when you ask them to amplify themselves, you just get more noise.
Nate Berkopec tweet media
English
36
22
358
24.6K
Noah Hayes
Noah Hayes@ninjoah·
@MTabarrok I like how the fireplace mantel is literally George and a Bush.
English
0
0
0
632
Maxwell Tabarrok
Maxwell Tabarrok@MTabarrok·
Literally looks like Trump's oval office is the 3rd or 4th step on the "make it EVEN GOLDER" back-and-forth with image models
Maxwell Tabarrok tweet mediaMaxwell Tabarrok tweet media
English
172
56
1.9K
10.1M
Noah Hayes retweetledi
Tim Sweeney
Tim Sweeney@TimSweeneyEpic·
Best wishes to everyone in Iran. It's one of the most ancient centers of civilization and scholarship, full of generous people living without freedom while filling universities and companies around the world with talent. Iranians deserved better than this past 73 years.
English
178
294
4.3K
208.7K
Noah Hayes retweetledi
Saurabh Srivastava
Saurabh Srivastava@_saurabh·
Code is an amazing petri dish for measuring and building intelligence. We released Rnj-1 yesterday; a 8B code & science model that you can run offline on your laptop. It got 20.8% (!) on the SWEBench software engineering benchmark. For context: a) that beats all comparable 8B open models by 10x, and b) punches way above its weight. It beats Google's Gemini 2.0 Flash and Alibaba's Qwen2.5 Coder 32B. This performance is close to OpenAI’s GPT-4o and their 120B open model despite being much smaller. This is where Essential AI is starting. Tons to come over the next year. Key to Rnj-1 was extremely strong pre-training (data, optimizers, mixing, infra, evals). Tons of distributed data work. Daily ablations on 0.2–2Bs to find the many hyperparams. New data distributions that teach the model code semantics beyond text, including leveraging PL ideas. Does code & science intelligence translate to other domains? Try it, tune it, and let us know. We're as excited to hear where it fails as much as what it does well.
Saurabh Srivastava tweet media
English
1
1
16
1.1K
Noah Hayes
Noah Hayes@ninjoah·
@its_The_Dr Bro these are 2 completely separate videos stitched together with a made up story.
English
0
0
1
56
Noah Hayes
Noah Hayes@ninjoah·
@CNviolations The lock should be foot operated so you don’t have to stick your hand in the trash
English
0
0
0
24
Ryan
Ryan@GatorRyan·
Starting Saturday
Ryan tweet media
English
49
1.9K
27.4K
2.1M
Noah Hayes retweetledi
WarrenBuffering
WarrenBuffering@WarrenInTheBuff·
Sorry I missed your update, I mainly try to keep an eye on our: - Gmail - Microsoft Outlook - Slack - Microsoft Teams - Jira - Basecamp - Figma - Mail(dot)0 - {Internal Proprietary Encrypted Messaging Platform} - Google Chat - Confluence - ClickUp - Asana - Linear - Zoom - Google Meet - Webex - Tuple Chat - Chanty - Workplace Chat - Mattermost - Discord - Yammer - Around - RingCentral - Dialpad - Twilio - Vonage - WhatsApp - Trello - Monday(dot)com - Notion - Google Docs - Blue Jeans - GoToMeeting - Zoho Cliq - Bitrix24 - Shortcut - YouTrack - Phabricator - Intercom - Wrike - Smartsheet - Podio - Hive - Quip - Coda - Airtable - HubSpot - Salesforce Messages - Freshdesk - ZenDesk - Front Inbox - Help Scout - SMS - Phone Calls - Physical Letters in my Mailbox - Rocket(dot)chat - Element - Github Issues - Gitlab Issues - Bitbucket Issues - Glip - Flock - Lark - WeCom - Workplace by Meta - Redmine - Trac - Mavenlink - Celoxis - Freedcamp - Proofhub - Nulab - Taiga - Zoho Projects - Flow - Olark - Drift - Crisp - Kayako - LiveAgent - Tawk(dot)to - 8x8 - Ooma Office - Nextiva - Mitel MiCloud - Grasshopper - Bloomfire - Samepage - Hypercontext - Zoho Connect - Ryver Just hit me up there if you ever really really need to reach me fast
English
174
319
5K
223.1K
Sandy Petersen 🪔
Sandy Petersen 🪔@SandyofCthulhu·
One day I would like to see a logo redesign that made it look better instead of soulless and terrible. Can anyone give a counterexample?
Sandy Petersen 🪔 tweet mediaSandy Petersen 🪔 tweet mediaSandy Petersen 🪔 tweet media
English
3.8K
10.5K
317.2K
29.1M
Chris Frantz
Chris Frantz@frantzfries·
@ninjoah Yeah, that’s what it would take to git me to commit
English
1
0
3
86
Chris Frantz
Chris Frantz@frantzfries·
BIG NEWS! GitHub was named a Leader in AI Coding and Software Engineering Technologies! Wow!
Chris Frantz tweet media
English
6
0
41
4.1K
Dylan
Dylan@DylanMcD8·
A great example of why the Touch Bar was so good: when tapping the brightness control, you could change the built-in display or your connected monitor (Studio Display in this case). I will never forgive them for getting rid of it.
English
131
38
1.5K
177.1K
Tyler Shukert
Tyler Shukert@dshukertjr·
We revamped the design of our dashboard at Supabase.
Tyler Shukert tweet media
English
615
404
15.4K
708.7K
Jim Aylan
Jim Aylan@JimAylan·
@localghost SF doesn't look like that small park you conveniently chose out of the entire place for your picture.
English
1
0
5
179
Aaron Ng
Aaron Ng@localghost·
Tokyo is basically the opposite of SF. SF looks like the past but builds the future; Tokyo looks like the future but is rooted deeply in the past. A city that struggles to fit 1M vs. 40M with no problem at all.
Aaron Ng tweet mediaAaron Ng tweet media
English
3
0
57
3.2K
Noah Hayes
Noah Hayes@ninjoah·
@ericmigi The mistake here is not expanding TAM. These new services are intended only for existing AirBNB stays which eats into the already slim budgets of Airbnb travelers. What if services were useful beyond just travel—think ghost kitchens or workout training sessions hosted in homes.
English
0
0
0
115
Eric Migicovsky
Eric Migicovsky@ericmigi·
I guess this is what could happen when you have a maniacal focus on quality, craft and design
Eric Migicovsky tweet media
English
40
50
1.3K
204K