Frank 👋

🤠 Co-founder & CTO W21

[Frank 👋 @FrankLiceaHowdy]

@paulg Hardest lesson is realizing kids grow up to be their own person. Just another adult in the world. Theres only a slight influence that parents have.

[Frank 👋 @FrankLiceaHowdy]

Companies are frustrated that their teams are not picking up AI fast enough.

So they mandate tool usage, track adoption, and even fold AI into performance reviews.

Most of that pressure ends up on engineering teams.

But that is not where the real friction is. It is company culture.

What many managers miss is this: if you want AI first workflows, you usually have to loosen your grip on deadlines first. AI first engineering is not just “use Copilot more.”

It is a different way of building software, and it requires time spent on work that does not immediately ship features. Early on, it often makes teams slower, not faster.

That work looks like things most engineers already know are important but rarely get time for:

Automated test suites
Preview environments
Real integration setups
Linters and type checking
Fast, boring CI

These are the foundations of sustainable software engineering. They also compete directly with the pressure of “we need this next week” or “launch by the end of the month.”

Then leadership wonders why AI tools do not stick. Of course they do not. The culture is not permissive of experimentation. AI first work needs false starts, delays, and infrastructure investments that have to come before any visible product value shows up. When that space does not exist, engineers quietly fall back to old habits.

Now contrast that with engineering forward cultures. These are places where failed prototypes are normal, where slower but smarter approaches are encouraged, and where “playing with the tech” is part of the job, not something done at night or on weekends.

Those companies will win the AI engineering race. Not because they mandated tools.

But because they made it safe for engineers to learn, fail, and build the foundation first.

And the uncomfortable question for leaders is not “why is my team slow to adopt AI?”
It is “have I actually given them permission to?”

[Frank 👋 @FrankLiceaHowdy]

AI inference dropped from about $20 per million tokens in 2022 to roughly $0.07 today.

That kind of price drop usually means everyone is building the same thing at the same time. When that happens, prices fall and big capital bets stop paying off. We’ve seen it before with fiber networks and electricity.

Compute will keep getting cheaper. For CTOs, the real question is where leverage lives once it does. History says it’s not the infrastructure. It’s the workflows, integrations, trust, and switching costs layered on top.

[Frank 👋 @FrankLiceaHowdy]

After listening to the latest AI research discussion, I’m less interested in AGI timelines and more interested in org design.

This clicked for me listening to Lex Fridman Podcast #490 with Nathan Lambert and Sebastian Raschka.

They’re talking about today’s large language models. Things like OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini.

It’s already clear these systems outperform humans in important ways. They absorb and recombine massive amounts of information, move extremely fast, don’t get tired, and are very strong on short, well defined problems. You can see the impact inside real teams today.

The limits are just as clear.
They don’t really learn from experience. Memory degrades. Mistakes don’t reliably turn into lasting improvement. Over longer periods, plans drift and errors accumulate. Those limits come directly from how these models are built.

One thing I found notable is that the researchers themselves don’t seem very focused on classical AGI anymore. Not the idea of one system that can learn to drive cross country with minimal instruction and then switch over to doing original physics research. That just isn’t what’s driving the work.

What’s driving it is human leverage.

A smaller number of experienced people can now explore more ideas, ship faster, and cover more surface area than before. The bottleneck shifts from headcount to judgment.

Teams using this mainly to cut people will get short term wins. Teams using it to increase the reach of their best people will compound.

That gap is going to matter more than any AGI timeline.

[Frank 👋 @FrankLiceaHowdy]

@DjDeVries3006 @txkyoghxst @grok @JasonHo94813957 @ferllenaf @_oRyca_ @EndWokeness While the report noted that there were “no life-threatening injuries identified,” this referred to the absence of specific traumatic injuries like broken bones or internal bleeding. It did not imply that the physical restraint was not fatal.

[Frank 👋 @FrankLiceaHowdy]

AI is creating a permanent underclass of engineers.

There’s a lot of talk about AI taking over software engineering jobs, but that’s not exactly what's happening.

A recent study found that AI coding tools like GitHub Copilot are making software engineers more productive, with an average increase of 26% in completed tasks. This boost is more pronounced among junior engineers, who show higher adoption rates and significantly greater productivity gains compared to their senior counterparts.

Here’s the math: juniors are producing code faster than ever, but the study shows that senior engineers aren’t experiencing the same level of productivity boost. This imbalance can lead to a bottleneck where code is produced faster than it can be reviewed and integrated. Imagine a situation where juniors increase their output by 50%, but seniors only improve their review capacity by 25% because they have to double check AI generated code. Assuming that Senior engineers review all the work done by Jr. Engineers, then the result is a bottleneck—code piles up waiting for review, slowing down the workflow or forcing code through without proper checks. Neither option is ideal.

This dynamic creates a real problem. Juniors are leaning heavily on AI to produce more code, but they're not developing the deeper skills needed to advance into senior roles. Meanwhile, there aren’t enough senior engineers to handle the increased volume of code. This mismatch not only affects workflow dynamics but stalls the professional growth of junior engineers.

We're looking at the rise of two distinct groups:

AI Code Producers: Juniors who stay stuck as juniors because they rely on AI for most of their work and miss out on the experiences that build deeper expertise.

True Seniors: A shrinking group of engineers who can review and guide the AI-generated code safely into production for the long term.

So no, AI isn’t replacing software engineers. It's creating a divide. And the concerning part? We might end up with a permanent gap where one group is stuck in junior roles with high output but limited growth, while the other group becomes rarer and more overwhelmed with the responsibility of maintaining quality.

What do you think?

[Frank 👋 @FrankLiceaHowdy]

Something no one is talking about is how much AI coding is going to cost companies.

I just spent $5 of AI credits to make a Tetris game. That’s shitty. Now imagine the cost on a large enterprise codebase.

Here’s a new coding test: using AI, build a playable Tetris game and keep the cost under a dollar.

[Frank 👋 @FrankLiceaHowdy]

I even asked it to visualize the A* algorithm. Nailed it.

[Frank 👋 @FrankLiceaHowdy]

I just asked the new OpenAI o1 model to help me make the start of an RTS game. Damn it’s scary good. Clicking, pathfinding, collision detection, obstacles, vector normalization, frame elapsed time, it’s all there out of the box.

Everyone can be a game designer now.

[Frank 👋 @FrankLiceaHowdy]

Relying solely on Y Combinator for your company's success is a misconception.

Y Combinator is a valuable thinking tool, providing access to brilliant minds to work through problems. However, true accountability and insights must originate from within your own efforts.

Here are some tips to maximize your Y Combinator experience:
- Ownership is Key: Leverage Y Combinator as a tool, not a magic solution. Take ownership of your company's direction and success.
- Engage Actively: Make the most of the brilliant people assigned to you. Actively engage in discussions and seek their insights, but remember that the driving force should be your commitment.
- Set Clear Objectives: Define your goals and expectations from the Y Combinator experience. This clarity will guide your interactions and help you extract the most value.
- Proactive Problem-Solving: While Y Combinator provides a platform for collaborative problem-solving, be proactive in identifying and addressing challenges. Initiative plays a crucial role in success.
- Build Your Network: Beyond Y Combinator, cultivate a broader network. Connect with fellow founders, mentors, and industry experts to complement the insights gained within the program.

Remember, Y Combinator is a catalyst, not a guarantee. Your proactive approach and commitment will ultimately shape the trajectory of your company's success.

[Frank 👋 @FrankLiceaHowdy]

How do you grow into the founder or executive your company needs to succeed?

The fastest way is to learn from others, but it’s a tricky balance. First, you need to form a strong opinion about what's stopping your company, then align with subject matter experts (SMEs) to address challenges swiftly.

But working with an SME doesn't mean blindly accepting their views. Collaboration and adaptation are critical to incorporating their expertise into your business's context. After all, no one knows your business better than you do.

However, it’s tricky when it comes to engineering. For example, when founders delegate to agencies to build their product, it makes it difficult for founders to form strong opinions about the tech and this is dangerous.

I often speak with founders who have fallen out with their agency due to delayed product updates or excessive costs.

So if you’ve finally reached the limit with your agency (or any painful relationship) here are some things for you to try before you hit the nuclear option:

1) If your teammates are being used for other projects, ask to have them relocated to a different co-working space and ensure that the agency puts a stop to this practice.

2) If your agency only does well when a specific manager is involved, ensure they stay on your project full time.

3) Retaining top agency developers can be difficult. Make sure to know your engineer's take home pay and the agency's margin in order to provide them with the necessary compensation.

4) If you buy out, don't burn bridges. The agency helped find a dev who propelled your business forward, and you may need their help again in the future.

5) Acknowledge the costs the agency put into finding your dev: employer marketing, recruiting, technical vetting, local resources like offices, equipment, healthcare, local taxes, benefits, management, etc. By working together you can avoid burning bridges.

6) If you are successful in hiring the developer from your agency, make sure to budget for more than just their salary. Don't overlook the importance of cultural perks such as birthday gifts, care packages, company merchandise, team lunches, conference attendance, and ongoing education. It is crucial to form strong relationships with your colleagues to avoid being targeted by competing companies in the local market.

What other tips do you recommend?

[Frank 👋 @FrankLiceaHowdy]

If you're exploring the idea of expanding your team in Argentina or already have a remote setup there, it's essential to be in the know about recent developments:
In December, Javier Milei assumed office, ushering in the potential for significant shifts in monetary policies that directly influence hiring dynamics in the country. Shortly after his inauguration, Milei issued a DNU (Decree of Necessity and Urgency), marking a substantial change in Argentina.

Notably, he introduced a measure allowing agreements to be freely conducted in dollars, a practice previously restricted by law. Also, Milei stated that Argentina is in a phase of pre-dollarization and explicitly communicated that purchasing the blue dollar (historically considered illegal) would not be unlawful under his administration.

It's important to note that although he made these statements, the actual implementation of these measures is yet to be confirmed. These developments have significant implications for businesses and hiring strategies in Argentina, especially those overseeing remote teams. Staying abreast of these changes is imperative for anyone navigating the evolving business landscape in the country.

Now, considering these shifts, I'm thinking: How will they impact contractors, and will the local market start competing with the global one? Could Argentina cease to be a convenient option for nearshoring? Conversely, might complete openness to international negotiations prove highly advantageous for the talent market?

[Frank 👋 @FrankLiceaHowdy]

How do you get a software development team to ship on time?

At work we've helped build hundreds of software development teams and we've observed their various failure modes. Here are some tips based on what we've observed:

1) Respect the details: Designers and product managers subconsciously resist the effort of fully illustrating detailed interactions, leaving developers to fill in the gaps. If your developers have to fill in the design gaps, you can kiss all your deadlines goodbye.

- Consider this: if a designer spends 3 hours on a mockup, it could take the dev team 30 hours to build. So even if the designer spends 6 hours on more detailed mockups, it ends up saving time in the long run.

2) The curse of expertise: CEOs, product managers, and designers often assume their designs are obvious. This can lead to confusion and the need for costly redesigns during development, causing the team to miss its deadlines. That's actually the happy path. If your dev team is less mature, they'll implement the confusing designs as-is, leading to embarrassing usability bugs for customers to find! Always test designs with new users, no matter how simple the designs may seem.

3) Embrace waterfall planning: Developers are so eager to write code and designers are so eager to design, that the slightest degree of planning feels like it goes against agile methodology. Resist this response from your teams, if they claim you're asking for waterfall then you might be doing it right! A common mistake that teams make is not having a concrete plan for their sprints. They may think that they can be flexible and let the project evolve as they go, but this lack of structure often leads to missed deadlines. It's important to have a clear vision and all the necessary elements – stories, interactions, empty/error states, and other details – ready for the initial release in order to stay on track.

[Frank 👋 @FrankLiceaHowdy]

There isn’t necessarily a right path into entrepreneurship and building businesses. So if you feel like you’re not moving fast enough, or that maybe entrepreneurship is not for you, don’t let that discourage you.

[Frank 👋 @FrankLiceaHowdy]

The path to entrepreneurship isn’t always what YouTube celebrities show you, like the hustlers in high school who became multimillionaires by the time they were 18. To be honest, mine was to fail side projects, fail startups with friends, years of not doing side hustles, and just putting my head down trying to be a good employee, learning and investing in myself.

[Frank 👋 @FrankLiceaHowdy]

The payoff is that by the time you feel ready to dive into the pool, and really start your own business, you’ve built up a network people that know your work, your integrity, and these relationships become your future investors, clients, mentors, or introductions to things or other people you didn’t know you were that close to.

[Frank 👋 @FrankLiceaHowdy]

Stop being a hard worker and become a smart worker.

Being a smart worker means having the ability to understand which of your tasks are high leverage and lead to actual productivity.

[Frank 👋 @FrankLiceaHowdy]

Google, renowned for hiring top talent, has open roles in Latin America, including Go-To-Market Lead and Security Cloud Consultant. Building teams in Latin America offers real-time iteration advantages over offshore teams. Nearshore teams provide quick fixes appreciated by customers, fostering proximity and effective collaboration. The short flights, like 4.5 hrs from SF to Mexico City, and 5 hrs from Austin to Bogotá, boost face-to-face time crucial for startups initiating significant projects.

[Frank 👋 @FrankLiceaHowdy]

Nearshore staffing with Latin American talent has skyrocketed in the last three years, increasing by 286%. Out of a pool of more than a million developers in Mexico, Brazil, Argentina, and Colombia, the US makes up an impressive 63% share of employers that hire from this global talent market.