Shindi

Really cool framing, never saw this broken down this way

Really cool project, it looks really useful. We’re moving past manual prompt optimization and considering different options for tuning long horizon tasks. We will likely go with Horizons

Tbh this model is extremely bad. I tried a couple of our medical form examples and it couldn't find almost any of the fields.

Remember that Gaza has the most amputated children per capita in the world. Next time you see a child, you should remember this fact. because Palestinian children aren't this abstract concept, but real people just like the children in our own neighborhoods

I totally missed you can embed AI features directly into claude artifacts. I ran the first GPT store hackathon in NYC and I feel like people were very underwhelmed. What could you do in a custom GPT you couldn't with a shared chat? The integration felt very annoying. Claude I feel nails how easy it is to build + how easy it is to share

The best way imo to get early sales for your startup is to use your network. Look at people you directly know who have the problem. I feel like this should be enough to get your first customer or two. After that I'd look at warm introductions.

I've found that in today's world, people are fatigued with inbound email and LinkedIn and so even if you have something valuable, getting someone to listen can be hard. So I think warm intros help at least get to the demo stage.

You might not think you're super connected but via six degrees of separation, you should be able to get a dozen or so demos booked via people who know people in your ICP.

If that doesn't work for your first ten customers (or first customer if enterprise) then you have to seriously question if you're working on the right problem.

You mentioned offering deals or rewards for trying out your product. I recommend against rewarding people for just trying your product. Because if the pain is not so great that they are willing to try out the product on their own, you're probably working on the wrong problem.

This is super cool! Huge upgrade to my product screenshots. Wondering if you're offering this as a react component - something I can embbed with a lead magnet or on a site.

I'm pretty sure it's actually hair brained

If you spend two minutes googling, you'll find loads of tutorials. You're not lacking information, you're lacking the drive to do basic research

I left big tech to start a company. The timing worked out great in my case. I was already trying to leave and the layoffs gave me an "angel round" to be able to pursue ideas. The biggest surprise was the uncertainty and amount of hours you have to work at a 0-1 company.

Current idea is taking off. I'm solving a global problem that affects literally every company. We have early design partners but looking to team up with a cofounder before we get further along.

If you're a big tech worker who was laid off an actively exploring ideas and want to team up, hmu! Email + twitter in my bio

> when I learned it was common for a founding engineer to get like half a percent of the company, I was appalled. I'd rather just start my own startup

I don't think people appreciate how hard it is to go from 0-1. To validate a product/market, get initial customers, before raising, heavily derisks the business and 1% is fair. Sure, you could just start your own startup, but you'll have to go through the brutal glass eating process of doing exactly that, starting a company.

In this thread: a ton of people with european work standards

The best way is to make sure you ACCOMPLISH something. I did the same as you and actually got a massively better role after my startup failed. The reason was because I accomplished something that's hard to accomplish.

How about you get a client before leaving? You should be able to find someone to pay for what you're making. If you can't work an additional 10,15,20 hrs per week at a cushy FAANG job and get a client you probably shouldn't leave to work the brutal life of rejection building a new company entails.

> They promised me an onsite interview to simply meet the team and potentially secure an offer.

Was that your first tech onsite? The onsite is an interview you have to pass, you're not guaranteed an offer. Also why did you not bring your laptop to your tech interview lol

Honestly, sucks to fly out far and get rejected but that's part of the process my guy. Most onsites are 5 back to back interviews.

Wow this is simply not true, Microsoft Clarity is free forever and has better analytics. I recommend it to any founder or builder because it comes with free unlimited heatmap.

https://clarity.microsoft.com/pricing

It really depends on what you want to do. Research (Foundational model layer)? Build cool LLM apps for work or play (application layer)?

Honestly the best way to start is to play with the tools excessively. Prompting over and over and over again will teach you amazing skills and intuitions.

Seeking co-founder | B2B AI | Engineering cofounder | NYC Only

Heya, I'm looking for a cofounder who is interested in making LLM tools for businesses. We already have a paying customer paying significant $$$. I'm a senior engineer who is looking for other talented engineers. Ideal person is in NYC and is playing around with langchain or similar LLM tools. You can still be at your job while we experiment together, no need to quit.

LinkedIn: https://www.linkedin.com/in/saleh-hindi-55895144/

Email: saleh.hindi.one@gmail.com

If you're using a regex it's a smell that there is a better tool for what you need to do. A ton of if statements is a much more readable way of writing code than regex.

This is a good example of bad use of AI. If your developer wrote code and tried to open a PR without talking to anyone, without testing, without an observability plan, without even running the code to make sure there are no errors, that would be crazy!

Good AI systems will do all the above.

Sorry to plug, but if you're a developer interested in building on top of langchain and building similar tech, please email me (in my profile). I'm a senior developer looking to collab.

Open source does NOT equal safe, it's actually worse. If you release something that can wreak havoc and it's open source, it will ALWAYS be out there, no off switch.

Imagine we later discover that an open source LLM is way more powerful than we realize. For example GPT-3 is pretty powerful but it really hasn't been out a while. Imagine what we discover it can do in 3-5 years, without even accounting for more advanced models like GPT-4 which is already out. Imagine someone discovers some really powerful, dangerous capability years down the line.

People can't imagine what can go wrong with LLMs, but think about the recourse we have for bad behavior online now: arresting people, forcing legal action against individuals/companies, sanctions or financial repercussions. Notice these aren't technical barriers, these are social barriers. You can't do these things again language models!

This. If the idea of doing sales or marketing at a small startup turns you off, don't become an indie hacker because that is your main job - to promote what you've made.

If you are trying to find super high quality cofounders, you should go to startup events irl in NY or SF.

You worked for a company and then made something really similar? idk, if what the author did was right or wrong legally, but sounds like a dick move. Just build literally anything else.

Surprised mentioned GPT-J! Here is a colab link: https://colab.research.google.com/github/NielsRogge/Transfor...

Although you need a premium GPU. I admit it's not as good at zero shot or 1-shot as GPT-3 but if you provide examples, you can get as good of output. I feel like the team behind it needs better marketing.

> E.g. effective altruists credit the B&MGF with basically removing all possible "room for more funding" wrt. vaccinations in low-income countries

This is wrong. idk anyone who believes this

This is a low quality and nonsense article. ChatGPT-4 doesn't exist, neither does chatGPT3. GPT4 is also not a fine tuned version of GPT3.

I think you're right, my comment isnt to disagree with you. You can make pairing and constant communication work remotely. However you have to make it work so to speak.

Meaning, there is a little bit more friction. You have to harangue senior engs to pair, you have to be posting in slack more, you have to do all these little things. But when you're in person there is less friction and so it gets done more naturally.

Like if a task is taking longer than expected, sometimes its because it takes longer than expected. In remote world, you have to know when to ask for help. In person, you might be casually complaining about a task and someone who knows a better way can chime in and get you unstuck without you realizing you're stuck.

> * The testimonials use fake names + stock art

Wtf this is true. Reverse google image search the testimonials and you'll see the same. This blows my mind how sketchy it is. Just get your friends to put testimonials. If you have 3 users that you begged to use your product, get them to review. Fake reviews is beyond scummy.

That said, anyone know what template they use for their landing page? It's nice.

Agreed. There is quickly going to be a bigger chasm between the best devs and average/bad devs bc of LLMs.

GPT-J feels much worse than GPT Davinci on the lack of instruct capability. My understanding is that it was not trained on instruction-completion data as much. It's great for text classification, summarization, and other tasks.

However, I also feel like it hallucinates/gives wrong outputs more. TBH i love what Eleuther is doing and GPT-J has its merits, but it feels not ready for primetime. I'd think of it as most similar to GPT Curie.