GPT-5 just thought for 86 minutes
& asked me 40 follow up / confirmation questions
& still hasn’t managed to answer my question.
@sama can I have the old models back please?
Prompt: make a debt analysis on a per person basis - government debt - credit card - mortgage - other debt
for the countries US, Greece, China, Russia, Brazil, Germany and Switzerland
Launching Nayla 2.0!
Your personal AI worker that combines a web browser & your work apps into one standalone product.
This combination enables your 500 open tabs to talk to each other unlocking the following capabilities:
- Talk to Nayla during video calls & execute commands
- Nayla infers tasks directly from your inbox & executes them proactively
- Extract data from the web into your work apps
(1/5) Super excited to announce Nayla - The AI Browser that works for you!
After months of sleepless nights and countless iterations, I’m thrilled to unveil my latest startup: @naylabrowser. Nayla isn’t just another browser — it’s a virtual worker that assists you at every step of your browsing journey and completes tasks independently.
Today, I’m thrilled to show you a sneak peek into the future of web browsing. Nayla is in early preview and still has many limitations, but I’m excited for you to explore its capabilities:
AI Browsing
Give Nayla any instruction and it will independently complete the task for you—making your online experience smoother than ever.
Voice Assistant
Nayla can talk to you, allowing natural, intuitive communication similar to chatting with a trusted colleague.
Memory System
By creating a digital twin of all the knowledge you gather—from websites to emails—Nayla retains all contextual information, ensuring you always have the right data for any future task or question.
Personalized Search
Unlike traditional search engines that often miss the mark, Nayla’s search engine is designed to deliver results that are tailored to your ongoing tasks.
Context-aware chats
With all the context of your open tabs and ongoing projects, Nayla’s chat experience is much more personalized and effective than standard chatbots.
These are just a few of the capabilities Nayla offers today—and there’s much more on the horizon!
If you’d like to be among the first to try Nayla, leave a comment below and I’ll personally guide you through our early preview. 🧵
(4/5) Context-aware chats
With all the context of your open tabs and ongoing projects, Nayla’s chat experience is much more personalized and effective than standard chatbots.
Next week I am 4 years sober from alcohol.
I always like to be the first to do something and I been using SONIA - AI Therapy.
For #WorldMentalHealthDay I reached out to @Sonia_Health to cover the subscriptions for anyone who signs up today.
https://t.co/cN57hm6r0Q
Help me help you. 🙏🏾🫶🏾
Just landed in San Francisco!
Exploring the consumer AI space and looking to connect with like-minded folks.
Also attending a couple events next weeks:
- YC Reunion
- SF Tech Week
- Chromium BlinkOn conference
In town until October 11th — shoot me a DM and let’s grab coffee!
I would like to express my deepest gratitude to my colleagues in the College, Commission services, MEPs, Member States, and my team.
Together, we have worked tirelessly to advance an ambitious EU agenda.
It has been an honour & privilege to serve the common European interest🇪🇺
Flike has been acquired! 🚀
After three years of incredible hard work and innovation, @flikeapp has now become a part of @firstquadrant. This acquisition aligns perfectly with our shared mission to empower tech companies to achieve sustained, exponential growth through AI.
Lots of confusion about what a world model is. Here is my definition:
Given:
- an observation x(t)
- a previous estimate of the state of the world s(t)
- an action proposal a(t)
- a latent variable proposal z(t)
A world model computes:
- representation: h(t) = Enc(x(t))
- prediction: s(t+1) = Pred( h(t), s(t), z(t), a(t) )
Where
- Enc() is an encoder (a trainable deterministic function, e.g. a neural net)
- Pred() is a hidden state predictor (also a trainable deterministic function).
- the latent variable z(t) represents the unknown information that would allow us to predict exactly what happens. It must be sampled from a distribution or or varied over a set. It parameterizes the set (or distribution) of plausible predictions.
The trick is to train the entire thing from observation triplets (x(t),a(t),x(t+1)) while preventing the Encoder from collapsing to a trivial solution on which it ignores the input.
Auto-regressive generative models (such as LLMs) are a simplified special case in which
1. the Encoder is the identity function: h(t) = x(t),
2. the state is a window of past inputs
3. there is no action variable a(t)
4. x(t) is discrete
5. the Predictor computes a distribution over outcomes for x(t+1) and uses the latent z(t) to select one value from that distribution.
The equations reduce to:
s(t) = [x(t),x(t-1),...x(t-k)]
x(t+1) = Pred( s(t), z(t), a(t) )
There is no collapse issue in that case.