hoa nguyen le

CEO @Dutycast - AI agents for Global Trade

[hoa nguyen le @lehoaams]

Hope to See you in 5 years when @dutycast we go public @LSEplc 🤞

[London Stock Exchange @LSEplc]

We were delighted to partner with Tech Nordic Advocates on their Shape the Future – Scaling Diverse and Inclusive Tech Businesses Internationally event yesterday, followed by the closing of the markets.

https://t.co/qAnMFzNQSo

[hoa nguyen le @lehoaams]

With the recent US Presidential Election signaling significant shifts in trade policies, businesses are entering uncharted territory.
We're launching #AI Trade #Agent - a coworker to save #lawyers and #trade professions days of research work and keep up-to-date with changes.

[Min Choi @minchoi]

It's only been a day since Anthropic dropped Model Context Protocol (MCP) with Claude, like universal plug for AI.

Now AI connects to tools and data insanely easier.

And people can't stop getting creative with it to do your work.

10 wild examples:

[Startup Archive @StartupArchive_]

Peter Thiel on how to identify great talent

Peter Thiel is famous for discovering undervalued talent throughout his career. The most well-known example is PayPal, where he recruited people who went on to start companies like YouTube, LinkedIn, Yelp, Affirm, and Yammer. But he also recruited Alex Karp, his former classmate at Stanford Law School, to co-found Palantir with him 2003.

Tyler Cowen asks Thiel what traits he looks for in undiscovered talent that everyone else overlooks. Thiel replies:

“It’s very difficult to reduce it to any single trait. A lot of what you’re looking for are these almost zen-like opposites.”

He gives a few examples:

“You want people who are both really stubborn and really open-minded. That’s a little contradictory. You want people who are sort of really idiosyncratic and really different, but then who can work well together in teams… If you focus too much on one end of it, you tend to get it completely wrong… [I like to look for] these combinations of unusual traits.”

Essentially he looks for idiosyncratic people with combinations of unusual traits that counterbalance those idiosyncrasies.

Video source: @mercatus (2015)

[hoa nguyen le @lehoaams]

How can this be sold for $1.5m but tech startups struggle to raise fund? 😂

[hoa nguyen le @lehoaams]

Is this too late to buy @Tesla stock?

[hoa nguyen le @lehoaams]

A student chat with AI as a therapist “@claudeai101 gave me a pat on the back like a therapist. @ChatGPTapp ranted with me like a homie. @Gemini lectured me on respect.”
https://t.co/OyAkmU8Vqg

[hoa nguyen le @lehoaams]

@rubenhassid Brilliant!

[hoa nguyen le @lehoaams]

When I started my first startup project, we were afraid people might steal our idea 🤣.
So now when investor asked about NDA, I just assumed he/she is not actively investing…. True or false?

[Andrew Ng @AndrewYNg]

Large language models (LLMs) are typically optimized to answer peoples’ questions. But there is a trend toward models also being optimized to fit into agentic workflows. This will give a huge boost to agentic performance!

Following ChatGPT’s breakaway success at answering questions, a lot of LLM development focused on providing a good consumer experience. So LLMs were tuned to answer questions (“Why did Shakespeare write Macbeth?”) or follow human-provided instructions (“Explain why Shakespeare wrote Macbeth”). A large fraction of the datasets for instruction tuning guide models to provide more helpful responses to human-written questions and instructions of the sort one might ask a consumer-facing LLM like those offered by the web interfaces of ChatGPT, Claude, or Gemini.

But agentic workloads call on different behaviors. Rather than directly generating responses for consumers, AI software may use a model in part of an iterative workflow to reflect on its own output, use tools, write plans, and collaborate in a multi-agent setting. Major model makers are increasingly optimizing models to be used in AI agents as well.

Take tool use (or function calling). If an LLM is asked about the current weather, it won’t be able to derive the information needed from its training data. Instead, it might generate a request for an API call to get that information. Even before GPT-4 natively supported function calls, application developers were already using LLMs to generate function calls, but by writing more complex prompts (such as variations of ReAct prompts) that tell the LLM what functions are available and then have the LLM generate a string that a separate software routine parses (perhaps with regular expressions) to figure out if it wants to call a function.

Generating such calls became much more reliable after GPT-4 and then many other models natively supported function calling. Today, LLMs can decide to call functions to search for information for retrieval augmented generation (RAG), execute code, send emails, place orders online, and much more.

Recently, Anthropic released a version of its model that is capable of computer use, using mouse-clicks and keystrokes to operate a computer (usually a virtual machine). I’ve enjoyed playing with the demo. While other teams have been prompting LLMs to use computers to build a new generation of RPA (robotic process automation) applications, native support for computer use by a major LLM provider is a great step forward. This will help many developers!

As agentic workflows mature, here is what I am seeing:
- First, many developers are prompting LLMs to carry out the agentic behaviors they want. This allows for quick, rich exploration!
- In a much smaller number of cases, developers who are working on very valuable applications will fine-tune LLMs to carry out particular agentic functions more reliably. For example, even though many LLMs support function calling natively, they do so by taking as input a description of the functions available and then (hopefully) generating output tokens to request the right function call. For mission-critical applications where generating the right function call is important, fine-tuning a model for your application’s specific function calls significantly increases reliability. (But please avoid premature optimization! Today I still see too many teams fine-tuning when they should probably spend more time on prompting before they resort to this.)
- Finally, when a capability such as tool use or computer use appears valuable to many developers, major LLM providers are building these capabilities directly into their models. Even though OpenAI o1-preview’s advanced reasoning helps consumers, I expect that it will be even more useful for agentic reasoning and planning.

Most LLMs have been optimized for answering questions primarily to deliver a good consumer experience, and we’ve been able to “graft” them into complex agentic workflows to build valuable applications. The trend of LLMs built to support particular operations in agents natively will create a lot of lift for agentic performance. I’m confident that large agentic performance gains in this direction will be realized in the next few years.

[Original text: https://t.co/gginTyOgwe ]

[hoa nguyen le @lehoaams]

@HarryStebbings Right! why doesnt everybody start doing that already 😅

[hoa nguyen le @lehoaams]

@briannekimmel Impressive!

[Rihard Jarc @RihardJarc]

At the end of the day, we all feel the pain of selling something too early.

The reaction of +$30B net worth Masayoshi Son's selling of $NVDA feels relatable.

[hoa nguyen le @lehoaams]

"So many regulation changes that it is so easy to miss out on important updates. We have to help each other to notify industry org when discover important changes so that they can act prevent uncessary stop in import/export"
#international trade #import #export #Compliance

[Iron Prairie Ventures @ironprairievc]

📊 Supply Chain Investments: The Rising Star in VC Funding 🚀 Exciting news from the world of venture capital! A recent Logistics Viewpoints article reveals that supply chain investments are making waves:

[Ole Lehmann @itsolelehmann]

I'm German.

16 years ago, the EU and US economies were neck and neck.

Today, the US economy is 50% larger than the entire EU combined.

Here's the devastating truth behind Europe's ongoing economic suicide 🧵:

[Startup Archive @StartupArchive_]

Sam Altman and John Collison explain that every startup goes through "the dark times"

Even though Open AI was recently valued at $157 billion, the journey to this point wasn’t as smooth as you might think.

As Sam Altman puts it:

“No one is as honest about how bad the early days are as they should be because it’s so embarrassing in retrospect.”

He talks through just how unpromising everything looked in the early days:

“We were this extremely ragtag group of people. We were mocked by everyone serious in the field. And we didn’t have working technical progress. We had some little things that kind of worked, but it was deeply unclear how we were going to make AGI, and we were unbelievably outgunned by Deep Mind at the time. A lot of people were like ‘why are you doing this? Deep Mind is untouchable.’… so yeah, it was pretty hard but we just kept putting one foot in front of the other. We were constantly not able to find enough money or compute or people. But you just keep going and eventually something works.”

Stripe cofounder John Collison reiterates this:

“Every company that is going to get to some scale definitely goes through a period of the dark times. If I think back to Stripe’s dark times, there was this point just after launch where the endorphins had worn off, and we had a bunch of early people leave all at once. It was very draining. You know, you lose some of your own confidence.”

What kept the Stripe founders going—besides having an idea that they were really excited about—was momentum from serving customers. When you have customers using your product, you kind of have no choice but to continue serving them.

Sam echoes this point:

“Once you have customers that you have to serve, they really pull it out of you.”

Open AI didn’t have customers and was more focused on foundational research, but to keep their sanity, they tried to set up external touch points that faked the equivalent of having customer feedback.

Video source: @stripe (2023)

[Aadit Sheth @aaditsh]

🤯 McKinsey says AI will automate 30% of work hours by 2030.

There's more in their 76-page long report on generative AI.

Here are the 5 most important things you need to know:

[Brian Halligan @bhalligan]

All of our trillion dollar tech companies were started by first time founders. Does anyone else think the second time founder thing might be overrated?

[Jenny Fielding @jefielding]

Caught up with 3 high net worth individual investors last week who typically invest directly into startups and also into VC funds. All come from founder backgrounds. All mentioned that they are now only focused on classic Private Equity opportunities. Wow - what is happening?

[hoa nguyen le @lehoaams]

@harjtaggar We streamline international trade with AI powered Trade agent