i built an open source version of OpenAI Assistants API
it's compatible with any LLM & embeddings model and u can swap it out for OpenAI assistants API with just one line of code.
if you want this let me know 👇
Had an epic evening with 52 brilliant hackers! 🚀
Thanks to the Assistants API, we went from idea to execution in less than 90 minutes, a task that once took so many days!
Want to reduce your AI development costs by 1000x? Check out projects and our link below! 👇
Mistral 7b + OpenAI API = 🔥
just released an example of using Mistral 7b (one of the best open source LLM) with an open source version of OpenAI Assistants API
OpenAI Assistants API is truly amazing - it automates decision making in your product
⭐️ 👇
just added insights from
@AnthropicAI
about your brain health through
@neurosity
's brain-computer interface
wear the Crown, record your focus, add tags, get insights, get better, weekly
FYI: Neurosity Crown collect your brain electrical activity
cc
@_buildspace
👇
just shipped an
@ouraring
integration
@neurosity
+
@ouraring
+ AI = striking insights about what affects your sleep and focus level throughout the days
soon you will know what events affects your anxiety according to your brainwaves and HRV
👇
just launched an app that let you become a better version of yourself based on your brain data
get insights on your life, what does good to you, what's harmful based on your electrical brain activity from the amazing
@neurosity
wearable
👇
common techbros' mistake: thinking that funding or accelerator is the goal
don't waste your energy on investors, focus on creating something useful and surviving
investors will be knocking on your door soon enough
OpenAI Assistants API for any LLM - open source
one liner away from switching from openai
atm supports
@AnthropicAI
but could easily supports any open source LLM
github link 👇
i built an open source version of OpenAI Assistants API
it's compatible with any LLM & embeddings model and u can swap it out for OpenAI assistants API with just one line of code.
if you want this let me know 👇
🤖 use OpenAI Assistants API with your own models in your own infra
one liner to use open source LLMs with OpenAI Assistants API
0 data leaked to openai
link 👇
🧑💻 I'm organizing a hack night focused on the new OpenAI Assistants API
This API enables building AI apps 10 times faster
In less than two hours, we'll create an app using the Assistants API
I'll also show how to switch to open-source LLMs w a single line of code.
👇
introducing Coconut Combinator (unrelated to
@ycombinator
)
a 1-3 months exploration / exploitation marathon for ambitious, curious, and inviting individuals from a beautiful villa on an island in Thailand
included: massage, cook, beach, adventure, top-notch ideas
👇
most people think there are 3 space companies on earth, actually there are thousands now and it's nonlinear with
@SpaceX
cutting cost for bringing mass-to-orbit by thousands
death is maximum entropy,
the end state of every system in the universe. but here's the thing - the choices you make today can have a profound impact on how quickly that entropy takes hold.
think about it this way: every time you opt for comfort over challenge, every time you
planning is overrated, sometimes the best outcome come from just keep doing what you love
"you can't connect the dots looking forward; you can only connect them looking backward"
- Steve Jobs
i just created an app that saves my brain electrical activity from
@neurosity
to the permanent web
@ArweaveEco
the human brain is the most complex object in the universe, and it's unfortunate that we don't keep track of our daily experience
👇 code
LLM devs' biggest mistakes:
- one-shot user interface (instead of a human iterative process in the loop, e.g. Figma)
- too much, harmful abstractions (e.g. LangChain, bad hard-coded prompts that can't be modified)
- too little context
- no streaming
- under-use of LLMs (use
RAG is just the tip of the AI iceberg
✨ The new OpenAI Assistants API combines RAG, function calls, code interpretation - LLMs now perform complex calculations.
AI apps that use Assistants API are 10x better
If you're not using Assistants API you're lagging behind
are you in control of your attention?
introducing 🧠^🤖 = mediar, google maps for your brain health
own an
@Apple
watch or
@ouraring
?
mediar let you keep tracks of your brain health
quit autopilot, become focused, beat distractions, achieve flow state 👇
i'd pay $2000 for a device that wrap my head like astronaut suit at night and give these feats:
- 100% noise isolation
- best air possible
- best temperature
- comfortable enough to sleep with it
- less than 2 kg, easy to carry
i'm neither e/acc, nor ea, nor k/acc, nor w/e ...
men are desperate for prescriptions
true philosophy is forged from the ground up - without relying on authority
The future of the space industry in SF is bright. If you're in the city tonight swing by the
@Interval
or follow
@StepsLeaps
to be invited to the next one.
"Create a react snippet that implements gnosis safe with wagmi?"
Had so much fun playing with the Embedbase
#GPT4
playground, so added the ability to talk to multiple data sources!
This example is using
@wagmi_sh
and
@safe
@mayowaoshin
Seems like the only missing block to turn LLM into a 100% programming language is to combine with traditional CS algorithms, data structures and pure logic
It takes so much time & money to organize information across all the software we use.
Time to get rid of everything, including the browser, and switch to
#AI
-augmented
@obsdmd
, let the magic of
#AI
organize information automatically for you.
in the near future we'll have 10 LLMs on the client interacting with 10 other LLMs on the server, just as many models interact with each others in the brain
I'm still waiting for the real web 3: 2d or 3d infinite canvas powered by AI.
What we've learned from brain-computer science is that every brain has very different patterns.
That's why the next web will adapt to each brain and more closely resemble our evolutionary environment.
a few months ago i interrupted several years of zero carb diet:
there are too many inputs in the equation, ppl say go carn, go vegan, go x, y, fast, cold showers, take pills w/e
“The first principle is that you must not fool yourself and you are the easiest person to fool.”
UX is 10x more important in application layer AI product than in non-AI product
LLMs are a lever on human intelligence and if you fail on the fundamentals it has a nonlinear impact
Great UX:
-
@cursor_ai
-
@perplexity_ai
-
@RewindAI
Get latest
@OpenAI
#ChatGPT
model in
@obsdmd
No context switching, drop all your software, feed all your life context & get your brain supercharged by ChatGPT
What's the best equivalent of a smartphone but without screen, hackable (e.g. Linux), discreet, durable (6h hours), internet-ready, end-to-end?
Basically a server in your pocket?
Or if possible as an implant :P
I'm a bit skeptical about "Scaling Reinforcement Learning from Human Feedback with AI Feedback"
The point of RLHF is to align AI to human. If you use an AI to align a new AI there is a higher risk of ending up with a paperclip maximizer, a disobedient AI
I’ve been using embeddings with LLM since 2020 in production and yes, embeddings are overkill
Just using full text search and generating the queries with LLM will remove a ton of infra headaches
If it’s too slow, use a small LLM for it. I.e like speculative generation
new personal knowledge management system I'm experimenting (in parallel to
@obsdmd
daily note) using
@excalidraw
Basically infinite canvas on a specific topic, just adding a box daily
Anyone experimenting on daily PKM in infinite canvases?
I think very few people yet understand that LLM are a new programming language. What you can do in 5000 lines of C, you can do in 500 lines of JS, you can do in 50 lines of
#OpenAI
GPT3.
The herd mentality in AI engineering—jumping onto every new trend or GitHub repo—is like being a billiard ball: predictable and controlled. Where are the first-principles thinkers?
publicly sharing the biggest mistakes I've made in recent years while working on my own startup:
- not being focused enough
- being too focused on the solution rather than the problem/user
- spending too much time building and not enough time selling
- letting myself be guided
📞 Open source function calling
Let your LLM do the right stuff in your product automatically - without being stuck in openai
OpenAI Assistants API but with any LLM
👇
OpenAI Assistants API but 75% cheaper, 23x faster, 100% data private
I'm working on a benchmark between an open source version OpenAI Assistants API vs the real one
Planning to compare
@AnthropicAI
vs
@MistralAI
vs
@OpenAI
for different tools, domains, etc.
👇 Want this?
working on an ai bitcoin advisor
it's to showcase an example using openai assistants api
built an open source version of this api
wanna contribute to this api? 👇
23 times cheaper, 75% quicker, totally data-private everyone can access it using the same api and sdk openai uses
After writing over 600,000 words on
@obsdmd
, I realized that my human memory is so limited and that this kind of software is on a good direction to expand our brain.
Rust is gaining popularity because learning to code with AI is easier than ever. Although Rust is more verbose than Python or TS, it saves time by catching errors at compilation, preventing crashes in user hands
Also, we’re moving towards Software 3.0, where most code is English
five years ago I trained AIs that would compete against each other for survival
just for fun
it was using similar algorithm than ChatGPT, reinforcement learning m
herbivorous vs carnivorous
carnivorous species seems always winner in history
Big congrats to
@anaprosina
, selected as one of
@payloadspace
's 30 under 30 Pioneers!
From 130+ nominees, stood out for her innovation and contributions to the space industry, evaluated by a panel of esteemed judges.
A star in the making! 🚀✨
OpenAI Assistant API is just one step closer to pure software 3.0
Programming in English only
At the moment we have to rely on software 1.0 hacks to do plumbing into LLMs context and letting it interact with the client from the server
rough summary of new openai abstractions
basically it's software 1.0 layers on top of software 3.0
I'm eager to see someone creating an open source nextjs api to replicate this in <200 lines of code