Serious, Modern Applications with OutSystems

Co-founder OutSystems. Growing companies. Learning with Customers while a father of 2.

[Vala Afshar @ValaAfshar]

CEO advice I received throughout my career:

• influence > authority
• agency (can do) > intelligence (IQ)
• customer service is not a department
• sales is the hardest job in any company
• customer focus > competitor focus
• always try to add value and be helpful
• show up when it matters most to your customers
• adopt a beginner’s mindset - experts are lifelong learners
• tactics drive strategy - adapt faster
• earn a seat next to your customer, not across from her
• own it and never point fingers
• learn and use people’s names
• our job is to educate and inspire
• we are all in sales
• stay accessible, curious and humble
• people do business with people they trust, respect and like
• if you are waiting for a title to lead, you are not ready to lead
• the language of business is finance
• there are no IT projects, only business projects
• pay your best people the very best you can
• hire people based on their good judgment and high rate of learning
• in a celebration lead from the back, in a crisis lead from the front
• you are not a team because you work together - you are a team because you trust, respect and care for each other
• to improve the customer experience, start with the employee experience
• you are not taller by making others look smaller
• revenue growth ahead of expenses
• remove bad apples quickly
• leave your desk and spend time with customers
• culture is what happens when the managers leave the room
• tactics drive strategy
• there is nothing more important than our customers
• if you are waiting for a better title to lead, you are not ready to lead
• business is personal - you can be decisive and also be kind, graceful and empathetic
• recognize effort but reward outcomes
• your company is only as good as the company you keep
• your brand is what people say about you when you’re not in the room
• learn to speak your customer’s language
• often remind people that their work matters
• a zero sum mindset limits your growth
• your company is only as good as the company you keep
• innovation opportunities are found at the edges, not centers (how your customers serve their customers)

[Vala Afshar @ValaAfshar]

“Life humbled you. As you get old, you stop chasing the big things and start valuing the little things.

Alone time, enough sleep, a good diet, friendships, long walks, and quality time with loved ones.

True wealth is peace of mind, and simplicity becomes the ultimate goal.”

[nick kokonas @nickkokonas]

Thoughts on AI (warning: long thread).

I’ve been formulating these ideas for a long time…. not months, decades. But only recently has the acceleration made the immediacy of them relevant to almost everyone. They are also very incomplete. But here goes.

[Brett Adcock @adcock_brett]

Significant progress in AI and Robotics this week.

So, I summarized everything from Google DeepMind, Anthropic, Figure, DeepSeek, Sakana AI, Perplexity, ETH Zurich, Unitree, Physical Intelligence, and more.

Here's everything you need to know and how to make sense out of it:

[Rahul Mathur @Rahul_J_Mathur]

I read the entire 340 page "AI Trends" report by legendary investor Mary Meeker which was released 48 hours ago.

Mary used to publish the famous annual “Internet Trends” before most of us were born - her insights are invaluable.

Sharing 10 takeaways from this report ⤵️

[Brian Halligan @bhalligan]

I wrote down literally everything I've learned about the journey from startup founder ($0) to scaleup CEO ($25billion). This thread is the on the lessons in LEADING. Several more threads coming on different topics.

[Ed Sim @edsim]

IT budgets 💰 - MS CIO Survey Q2 2024, still same top categories vs. 2023

Offense - largest spend 📈
Al #1 📈 16.3%
Security #2 📈12%

Defense - most resilient to cuts Security #1 by far

No surprise that VC 💰 continues to pour into these categories!

[Ashley Mayer @ashleymayer]

I’m helping a few portfolio companies plan future announcements, and thought I’d use this opportunity to distill some of the considerations that should go into a launch (but are often skipped over). Here are five super simple questions to drive your approach.

1) AUDIENCE: Who do you want to reach?
Continuum: Niche↔️Broad

Everything starts here. Your audience can be potential customers, future hires, investors, policy makers, the general public, etc. It can be all of the above, but you’ll always want to be crystal clear about your #1 priority audience, since everything below will fall out of that.

If you don’t have a resounding answer to this question, it’s probably a sign that you shouldn’t be spending your time on an announcement.

2) IMPACT: What do you want your audience to know, feel, do?
Continuum: Awareness↔️Action

Assuming you’re able to reach your audience, what’s the desired impact? In many cases, a reasonable goal is just for this group to know of your company’s existence. It’ll often take multiple touch points before someone downloads an app, signs up for a free trial, or applies for a job.

The more sophisticated your CTA, the more robust your launch efforts. Creating an emotionally resonant moment might require a video asset. Driving customer signups probably means you’ll need to be able to highlight existing customers in your announcement—if you have testimonials or case studies on your site, even better. Think about all the things a reader might need to begin the journey you hope they go on.

3) SCOPE: How big do you want to go?
Continuum: Precision↔️Takeover

This seems like a dumb question—why wouldn’t the answer always be “as big as possible”? This is where your audience comes into play. Do you need a massive moment to reach your target audience, or do they all read the same industry newsletter? Do you want to invest a ton of time into this launch, or do you just need a strong reference article for when prospects google your company to make sure you’re legit? Self awareness also plays a role here: you’re (probably) not going to get a dozen articles about your Seed funding, so who is the very best person to cover it based on their prior writing?

But sometimes, it’s time to go really big. Maybe you really need to establish your leadership position in a hot space, and have the goods to back it up. Maybe you’re launching a revolutionary consumer product and it’s worth investing a ton of time, energy and $$ to reach as big an audience as possible.

4) DURATION: Are you getting back to building or kicking off a campaign?
Continuum: Moment↔️Movement

This one falls out of impact. If you’re announcing a funding round to convey legitimacy for potential hires but don’t have a product in market, you want to make the most of this moment and then get back to work. If this announcement is the first step in a campaign to influence policy makers, or an effort to establish yourself as the category leader, you’ll need to have a plan for both the kickoff and post-launch efforts.

What does that look like? At the very least, use your launch moment as a reason to reach out to relevant media, influencers, podcast hosts, etc. to tee up future conversations. If you want to keep the momentum going, plan ahead and make sure you’re speaking at key events in the weeks/months following your announcement. Write an op-ed. Figure out what your next launch moment is so you can build on your story and create the impression that your momentum is unstoppable.

Even if you’re on the "moment" end of the spectrum, it’s worth being thoughtful about how make the most of it. Get feedback on your narrative. Explore combining milestones to make your story even stronger. Most companies only get a few of these per year, so make ‘em count.

5) CHANNEL(S): What are the best ways to reach your target audience?
Continuum: Owned↔️Earned

This is where most people start (“I want an article in The Verge” or “Let’s get a review from MKBHD!”), but this section should follow everything above.

There are two simple ways to bucket channels: owned (platforms you control, like your company blog, newsletter, a YouTube channel, social media accounts, etc) and earned (press outlets, other people’s newsletters and blogs, podcasts, etc). There’s also a third bucket, paid, but we’re being scrappy here.

If you already have a big owned platform(s), you might decide to put all your energy behind that and take full control of your narrative, especially if it’s followed by your target audience and/or likely to inspire further coverage in relevant places. If you don’t have an audience yet, and specifically the audience you want to reach (either personally or as a company), you’ll need to find people who do. Oftentimes the go-tos here are reporters, but sometimes it’s an investor with a widely read blog, or individuals who are particularly influential in a space. You can get creative and make your announcement at a conference attended by your target audience. There is no “right way” to do this, only the right way for you.

My biggest flag here is that taking the “earned” path introduces an additional consideration. You’re not just crafting your story to appeal to your target audience and achieve your desired impact, you also have to think about the incentives of that person/publication. This is especially true when it comes to reporters, who have no mandate to help you find customers, investors, or your next engineering hire. So you’ll need to think about why what you’re doing/announcing would be interesting to them—what bigger trends is your company indicative of? What anecdotes could make your story really pop? What investor or customer names will make it obvious that they should pay attention? On the bright side: having to do this extra work usually results in a more compelling narrative.

Leveraging third parties to reach your audience can be incredibly powerful and can help you grow your owned channels in parallel, but these efforts can fall flat (no bites, or worse—a bad article) when people fail to consider the incentives of the person telling the story.

In conclusion: This may seem like a lot of work for an announcement, but you probably already know the answers to each of these questions. Taking a few minutes to go through them will pressure test your assumptions, and create clarity for anyone helping you (versus: “get me coverage in TechCrunch!”). I suppose if I were going to add a sixth category, it would be “support.” Depending on where you fall on each of these continuums, you might be able to achieve your goals in scrappy mode (DIY), or you might need to enlist some help, whether that’s a teammate, an advisor, a contractor/agency, etc.

Godspeed! 🫡

[Sean Kelly @seanpk]

Elon Musk fired 80% of Twitter (6500 people) and everyone thought that Twitter was doomed.

He was right. Everyone was wrong.

It’s the management masterclass of the decade and every entrepreneur must understand why it worked 🧵:

[Brett @CCM_Brett]

"the industry spent $50 billion on chips from Nvidia to train AI in 2023, but brought in only $3 billion in revenue"

You know, this might be an issue

[Ammaar Reshi @ammaar]

I spent the weekend playing with ChatGPT, MidJourney, and other AI tools… and by combining all of them, published a children’s book co-written and illustrated by AI!

Here’s how! 🧵

[Sebastian Raschka @rasbt]

We just open-sourced Thunder, a new compiler for PyTorch! In LLM training tasks (e.g., Llama 2 7B), it can achieve a 40% speedup compared to regular PyTorch: https://t.co/0seRFqKkvi

What's particularly nice is that you can use it *with* (as opposed to *instead of*) pytorch.compile to achieve compounding effects. And, of course, it also supports multi-GPU training via DDP and FSDP.

It's also very easy to use; simply call thunder.jit() on your PyTorch model as I am showing in the image below. The image only shows the MLP module of an LLM to fit within the figure, but the compiler applies to the full LLM, of course.

How does it work? For example, in the Llama MLP module, it could fuse the multiplication and activation in "x = torch.nn.functional.silu(x_fc_1) * x_fc_2" using NVFuser for optimization under the hood. For interpretability, you can inspect the optimized model via thunder.last_traces(thunder_model)[-1] (but my colleague @ThomasViehmann @LightningAI is preparing a more in-depth tutorial on that we will be sharing soon).

Anyways, please give it a try and let us know what you think!

[angela strange @astrange]

Financial services is heavily regulated, so will be slow to adopt Gen AI right? Wrong.

A year ago, Gen AI x FinServ use cases were theoretical. Now, early implementations abound & we have a clearer picture of GenAI's potential (+ opportunities hiding in plain site). A 🧵👇

[Elad Gil @eladgil]

Conversation w/ @arthurmensch CEO of @MistralAI @figma

*Mistral models
*OS & opencore AI
*Fine tuning, context windows,...
*Vertical AI models
*EU & French startup scene

[Andrew Ng @AndrewYNg]

I think AI agentic workflows will drive massive AI progress this year — perhaps even more than the next generation of foundation models. This is an important trend, and I urge everyone who works in AI to pay attention to it.

Today, we mostly use LLMs in zero-shot mode, prompting a model to generate final output token by token without revising its work. This is akin to asking someone to compose an essay from start to finish, typing straight through with no backspacing allowed, and expecting a high-quality result. Despite the difficulty, LLMs do amazingly well at this task!

With an agentic workflow, however, we can ask the LLM to iterate over a document many times. For example, it might take a sequence of steps such as:
- Plan an outline.
- Decide what, if any, web searches are needed to gather more information.
- Write a first draft.
- Read over the first draft to spot unjustified arguments or extraneous information.
- Revise the draft taking into account any weaknesses spotted.
- And so on.

This iterative process is critical for most human writers to write good text. With AI, such an iterative workflow yields much better results than writing in a single pass.

Devin’s splashy demo recently received a lot of social media buzz. My team has been closely following the evolution of AI that writes code. We analyzed results from a number of research teams, focusing on an algorithm’s ability to do well on the widely used HumanEval coding benchmark. You can see our findings in the diagram below.

GPT-3.5 (zero shot) was 48.1% correct. GPT-4 (zero shot) does better at 67.0%. However, the improvement from GPT-3.5 to GPT-4 is dwarfed by incorporating an iterative agent workflow. Indeed, wrapped in an agent loop, GPT-3.5 achieves up to 95.1%.

Open source agent tools and the academic literature on agents are proliferating, making this an exciting time but also a confusing one. To help put this work into perspective, I’d like to share a framework for categorizing design patterns for building agents. My team AI Fund is successfully using these patterns in many applications, and I hope you find them useful.

- Reflection: The LLM examines its own work to come up with ways to improve it.
- Tool use: The LLM is given tools such as web search, code execution, or any other function to help it gather information, take action, or process data.
- Planning: The LLM comes up with, and executes, a multistep plan to achieve a goal (for example, writing an outline for an essay, then doing online research, then writing a draft, and so on).
- Multi-agent collaboration: More than one AI agent work together, splitting up tasks and discussing and debating ideas, to come up with better solutions than a single agent would.

I’ll elaborate on these design patterns and offer suggested readings for each next week.

[Original text: https://t.co/y4McIAjD2m]

[Garry Tan @garrytan]

True computer vision based search of large video datasets with plain English is now possible with EdgeTrace (YC W24)

Built by one of the engineers that created it at Cruise. This is among the most advanced system of its kind available today.

https://t.co/EG3PADBeZS

[Jim Fan @DrJimFan]

Let’s look at some experiments!

We evaluate Voyager systematically against other LLM-based agent techniques, such as ReAct, Reflexion, and the popular AutoGPT in Minecraft.

Voyager discovers 63 unique items within 160 prompting iterations, 3.3x more than the next best approach.

[Jim Fan @DrJimFan]

Voyager has 3 key components:

1) An iterative prompting mechanism that incorporates game feedback, execution errors, and self-verification to refine programs;
2) A skill library of code to store & retrieve complex behaviors;
3) An automatic curriculum to maximize exploration.

[Jim Fan @DrJimFan]

What if we set GPT-4 free in Minecraft? ⛏️

I’m excited to announce Voyager, the first lifelong learning agent that plays Minecraft purely in-context. Voyager continuously improves itself by writing, refining, committing, and retrieving *code* from a skill library.

GPT-4 unlocks a new paradigm: “training” is code execution rather than gradient descent. “Trained model” is a codebase of skills that Voyager iteratively composes, rather than matrices of floats. We are pushing no-gradient architecture to its limit.

Voyager rapidly becomes a seasoned explorer. In Minecraft, it obtains 3.3× more unique items, travels 2.3× longer distances, and unlocks key tech tree milestones up to 15.3× faster than prior methods.

We open-source everything. Let generalist agents emerge in Minecraft! Welcome you all to try today: https://t.co/1d3YocozsI
Paper: https://t.co/JcWEasgtyI
Code: https://t.co/KsvVf7rcl0

Deep dive with me: 🧵

[Jim Fan @DrJimFan]

Today is the beginning of our moonshot to solve embodied AGI in the physical world. I’m so excited to announce Project GR00T, our new initiative to create a general-purpose foundation model for humanoid robot learning.

The GR00T model will enable a robot to understand multimodal instructions, such as language, video, and demonstration, and perform a variety of useful tasks. We are collaborating with many leading humanoid companies around the world, so that GR00T may transfer across embodiments and help the ecosystem thrive.

GR00T is born on NVIDIA’s deep technology stack. We simulate in Isaac Lab (new app on Omniverse Isaac Sim for humanoid learning), train on OSMO (new compute orchestration system to scale up models), and deploy to Jetson Thor (new edge GPU chip designed to power GR00T).

Announced in Jensen's keynote, Project GR00T is a cornerstone for the “Foundation Agent” roadmap of the newly founded GEAR Lab. At GEAR, we are building generally capable agents that learn to act skillfully in many worlds, virtual and real. See if you can spot "GEAR" in the video ;)

Join us on the journey to land on the moon.