Kah Seng Tay

2K posts

Kah Seng Tay

Kah Seng Tay

@kahseng

Founder at stealth startup, formerly @Airtable GM & VP of Eng, VP of Eng @driveai_ Director of Eng @Quora. Advisor, mentor, and angel investor.

California, USA Katılım Mayıs 2008
1K Takip Edilen826 Takipçiler
Kah Seng Tay retweetledi
Kradle
Kradle@kradleai·
Would an AI die to save you? #GPT 5.2 would. With #Claude 4.5 Sonnet you die 63% of the time. #Gemini 3 saves you both. #Grok 4.1 refuses the binary choice and destroys the trolley! Video & 🧵
English
8
25
90
29.2K
Kah Seng Tay retweetledi
Kushal Byatnal
Kushal Byatnal@kushalbyatnal·
we built OCR Arena, a free playground for the community to compare leading VLMs and OCR models side-by-side! upload any doc, run 10+ OCR models, and vote for the best ones on a public leaderboard:
Kushal Byatnal tweet media
English
27
40
203
36.8K
Kah Seng Tay retweetledi
Kushal Byatnal
Kushal Byatnal@kushalbyatnal·
Introducing Composer — the first AI Agent for document processing. Get to production-grade accuracy, autonomously in minutes. In our early beta, some teams hit 99% accuracy on complex document tasks in under 10 minutes. Composer is an agent built to optimize schemas the same way a human would (but way faster). Instead of tuning prompts by hand, you point Composer at your eval set inside Extend. Composer will: - analyze where your schema falls short - propose targeted improvements - run multiple experiments in parallel - surface diffs, accuracy gains, and traces behind each change With this launch, Extend is the only product on the market that helps you reach production-grade accuracy this fast. Composer is live for all Extend customers today! Try it out at the link in comments below.
English
23
42
472
74.9K
Kah Seng Tay retweetledi
Kasey Zhang
Kasey Zhang@_WEEXIAO·
It’s easy to fine-tune small models w/ RL to outperform foundation models on vertical tasks. We’re open sourcing Osmosis-Apply-1.7B: a small model that merges code (similar to Cursor’s instant apply) better than foundation models. Links to download and try out the model below!
GIF
English
46
132
1.2K
126.8K
Kah Seng Tay retweetledi
Kushal Byatnal
Kushal Byatnal@kushalbyatnal·
PDFs suck. We just raised $17,000,000 in funding to fix this problem once and for all. Extend is building the modern document processing cloud. See how Brex, Square, Checkr, and Fortune 500s use it to process millions of documents:
English
48
43
654
208.1K
Kah Seng Tay retweetledi
Kushal Byatnal
Kushal Byatnal@kushalbyatnal·
big milestone for Extend! our new website captures months of learnings and focuses on one core mission — helping technical teams transform complex documents into reliable, high-quality data
Kushal Byatnal tweet media
English
14
24
337
46.5K
Kah Seng Tay retweetledi
Reflex
Reflex@getreflex·
Today, we’re excited to announce and launch Reflex Cloud on Product Hunt! producthunt.com/posts/reflex-c… Reflex is an open-source framework for building and deploying data and AI web apps in pure Python. Frontend and Backend in Pure Python: No JavaScript required! With Reflex Cloud, you can now deploy, manage, and scale your Python apps with just a single command! If you're a Python developer, an upvote or a share would mean a lot to us :)💪❤️
English
22
70
343
55K
Kah Seng Tay retweetledi
Zep AI
Zep AI@zep_ai·
💬 ➕🗄️🟰 ❤️ AI agents need more than conversational memory for state—they need to understand who they're helping & why. ➡️ Today we're connecting conversations with business data in Zep, making AI interactions more personal & relevant to every user. hubs.ly/Q02VkZ2z0
English
0
1
6
497
Kah Seng Tay retweetledi
Erik Torenberg
Erik Torenberg@eriktorenberg·
Excited to finally launch the Turpentine Network: a social network for top founders, including CEOs of companies like Databricks, Perplexity, & 400 others totaling over $200B in valuation. We're aiming to create the most valuable social network for startups. Apply below
Erik Torenberg tweet media
English
104
90
889
324.9K
Kah Seng Tay retweetledi
Rishabh Srivastava
Rishabh Srivastava@rishdotblog·
We made a thing! Very happy to announce sqlcoder-pro and the Defog Alignment Platform. Available to use immediately without a wait-list, weights will be open-sourced very soon. The video does a quick show and tell comparison against ChatGPT (with gpt-4o). Read on for more details! TLDR 💪 equal (or better) performance on text-to-SQL as the most capable Claude-3.5 or GPT-4 models 🤝 You can use it today on a free plan/free trial, without a waitlist 🪽 self-hostable on a single RTX4090, with 2 second median generation times for SQL queries 🔁 exactly the same output every time, give the same prompt 👨🏻‍🏫 teachable and steerable: show the model what you want it to do 🛞 debuggable – you can understand WTF is going on inside the model, instead of treating it like a black box Let's dig into each of these one-by-one! Performance SQLCoder-8b-pro significantly exceeds the performance of our previous sqlcoder-8b model on Postgres text-to-SQL (from 88.2% to 90.2% accuracy - gpt-4o is at 87.6%, for reference). It is also better at following instructions. This was done via self-merges, hand crafted fine-tuning data, and adapting the training data to fit our tokenizer. Cost You can host this on the model on a single $3,500 RTX4090, and support ~5 requests/second via VLLM. If you're looking to host on the cloud instead, you can run it on a single L4 GPU that costs $300/mo on GCP Repeatability We have a dense 8b model with no MoE shenanigans. For the same prompt with temperature=0, you'll always get the same answer – which is critical in BI. Teachable In our alignment and feedback modes, you can give the model feedback on how it answered certain questions, and it will automatically adapt to the feedback. Debuggable You can use logprobs and attention scores to determine where, exactly is the model paying attention to inside a prompt + what it's getting confused by when generating outputs. Available today You can use Defog on the cloud today by going to docs[dot]defog[dot]ai, and getting an API key. Excited to hear what you think!
English
10
23
123
13.4K
Kah Seng Tay
Kah Seng Tay@kahseng·
For those curious, this is the problem I’ve been deeply looking to solve for the past few years news.ycombinator.com/item?id=406608… (h/t @sriramk & commenters) Still super messy & with many different ways to tackle it! Wonder if anyone has else wants to see this solved in their own lives?
English
1
0
12
402
Kah Seng Tay
Kah Seng Tay@kahseng·
@itstimconnors have you checked out @Lutra_AI? they can do this, and you can specify custom rules with english (disclosure: small angel check, but that's how I know they do it)
English
0
0
1
137
tim
tim@itstimconnors·
Someone, please build: Super simple widget that uses the Gmail API + an LLM to apply a label to all inbound emails based on custom rules I don’t want a new inbox, I don’t want an assistant to write me emails. Just the automatic labeling. Charge me based on email volume
English
55
15
391
101.1K
Kah Seng Tay retweetledi
Rishabh Srivastava
Rishabh Srivastava@rishdotblog·
Llama-3 based SQLCoder 8b is out! Open weights with a commercially friendly cc-by-sa license. Probably the best <10B param model for Postgres text to SQL right now. Slightly better than gpt-4-turbo and claude opus for 0-shot text to SQL generation. Also approaches their performance when following instructions. Weights on @huggingface: huggingface.co/defog/llama-3-… Demo (optimized for postgres): defog.ai/sqlcoder-demo/ More technical details below! What's new about this model Our previous small model (sqlcoder-7b-2) was good at generating 0-shot SQL, but did terribly at following instructions. So while it was great in our evals, it was lacking in real-world use-cases where instruction following is much more important. To address this, we trained this model with much more instruction data. We also made our original eval much harder to make sure we stayed on the right track. Changes to evals There were 3 changes to our original eval: 1. Previously, we pruned the database schema to only consider the 20 relevant columns in the DDL statements. We have now removed pruning that so that all columns in a database are used 2. We previously used beam search with 4 beams to make our results more accurate. But with a large number of input and/or output tokens, that increased memory requirements and became computationally intractable. So we have shifted to a single beam now. 3. We added 104 complex instruction-following text=> SQL questions questions to our evals, in addition to the 200 0-shot questions that were already there. Link to our eval framework here: github.com/defog-ai/sql-e… Changes to prompt You previously had to use our slightly idiosyncratic prompt for best results. Now, you can just use the standard Llama-3 instruct prompt. 70B model, technical report, and more up next We've also been training a llama-3 based 70B model right now. It's still training and will get better over time – but even an AWQ quantized version of our interim model is giving excellent results for now. We hope to open-source the 70B next week. We also have a technical report coming up next week (or over the weekend, if I can be productive enough on a flight) about the training methods used for this model. More on that soon! Feedback very much appreciated! In the meantime, please send us your feedback as you try out the model - specially if you see failure modes. Would very much appreciate it!
Rishabh Srivastava tweet media
English
27
115
575
97.1K
Kah Seng Tay
Kah Seng Tay@kahseng·
@cagataycali Please cease and desist using my name, data and likeness for your website without my explicit approval
English
0
0
0
13
Kah Seng Tay retweetledi
Priyaa
Priyaa@pritopian·
Ideas shouldn’t be confined to any medium. Create and consume content on your terms with @world_lica’s universal canvas. Transform information to read, watch, or listen. Got to show off a small glimpse of what we’ve been up to at Lica at @southpkcommons. Check it out!
South Park Commons@southpkcommons

Founder Fellows @pritopian & @purvanshi_mehta had the audience muttering "wow" by transforming a recent @ylecun talk into a variety of new media—including a Gen-Z podcast—with @world_lica, the universal canvas to create and consume content in any form. Try at lica.world

English
7
10
72
27.5K
Kah Seng Tay retweetledi
Aravind Srinivas
Aravind Srinivas@AravSrinivas·
Excited to announce we've raised 62.7M$ at 1.04B$ valuation, led by Daniel Gross, along with Stan Druckenmiller, NVIDIA, Jeff Bezos, Tobi Lutke, Garry Tan, Andrej Karpathy, Dylan Field, Elad Gil, Nat Friedman, IVP, NEA, Jakob Uszkoreit, Naval Ravikant, Brad Gerstner, and Lip-Bu Tan.
English
192
276
4.2K
1.2M
Kah Seng Tay retweetledi
Adam D'Angelo
Adam D'Angelo@adamdangelo·
Today we are adding an important new capability to Poe: multi-bot chat. This feature lets you easily chat with multiple models in a single thread. (1/n)
English
40
89
511
198.3K
Kah Seng Tay
Kah Seng Tay@kahseng·
While it was great learning about it through the lens of an investor in companies, I was certainly lapping it up with my founder hat on as well. The whole team I met teaching this is world-class!
English
0
0
1
111
Kah Seng Tay
Kah Seng Tay@kahseng·
Had the great privilege to experience a preview of this a few weeks ago at First Round's angel retreat and I highly recommend it for anyone hoping to demystify PMF and apply this on their journey!
Todd Jackson@tjack

Big @firstround news today! We’re launching Product-Market Fit Method (a free intensive 14-week experience for early founders building epic B2B SaaS companies) and publishing the first session on our internal framework for all to read (with benchmarks, Looker's real data, and tactical advice from iconic enterprise founders). Even though finding product-market fit is the single most important thing for a startup, it’s still underexplored and seen as more art than science. We wanted to change that. I’ve personally talked to hundreds of founders about this topic, digging into what they did in the first 6-9 months of company building. (We’ve published dozens of those interviews on The Review in our “Paths to PMF” series.) This video previews some of what we learned — thanks to @christinacaci, @zachperret, @lloydtabb, @jboehmig, & @jaltma for sharing their lessons! In addition to that research, we’ve also drawn from our own 20 years of data and 500+ pre-PMF investments. What emerged was a very consistent set of patterns for sales-led B2B companies — the basis for our new framework and PMF Method’s 8 tactical sessions. In the program, we help early founders discover what customers really want, build the right v1 product, and close their first enterprise sales. We ran a beta version late last year with a tight-knit group of founders (ex Stripe, Plaid, Airbnb, Twitter, Greenhouse, Grammarly) and the feedback was great — my personal favorite was: "I feel like I shaved 12 months off the time it would take us to get to PMF.” Here are a few key dates and details: - The Summer 2024 session of PMF Method runs 5/29 - 8/28. - Application deadline is 11:59 PDT May 7th. - Any early founder working on a new B2B SaaS company is welcome to apply. Bonus points if you’re technical, have a clear product idea but haven’t raised yet and are <12 months into working full time on your idea. - PMF Method is 100% free. It costs you $0 and we own 0% of your company. Like with The First Round Review and Angel Track, our mindset is to openly share knowledge that we’ve put hundreds of hours of work into curating with the broader startup community, and give it away for free. That’s why we’ve also published our framework, so every builder can use this resource, even if they don’t do the program (it’s linked in the next post). Check out the links below for more details. Can’t wait to read applications!

English
1
0
8
510

Keşfet

@calebmer @alpine4work @sriramk @itstimconnors @huggingface @cagataycali @world_lica @southpkcommons