Lorenzo Pasqualis

This resonates so much. No, you can't do it all, especially not at a startup. Since COVID I see so many people try to balance unaided parenting with full-time careers, and no, I don't see anyone doing it well. My kids are now 17, 13 and 10 and just now, when they trek to most of their activities on their own, and do their HW on their own, and cook their own meal or snack, we have finally let go of the outside help ( and this is with both parents working from home for last 5 years, outside of peak COVID lockdown, we had always relied on a designated childminder outside of school hours). Yes, working remotely does mean that I have more time and flexibility in my life, it means I can walk my kids to school sometimes, or have breakfast with them instead of rushing to the office, it means I can go and see their midweek concert at school, or be there to grab them from school on a sick day in 10 mins. But it also means that on EVERY day basis we all need clear boundaries, work hours, dedicated spaces that are not continuously invaded by children, or feelings of guilt that our children are unattended. One thing that a lot of the arrangments lack is transparency, being explicit with ourselves, our partners, our managers and colleagues on when we are making trade-offs. What helps is verbalising our plans, putting explicit blocks on calendars for pickups, doctor's appointments and school activities, letting others know when you will have periods of diminished focus due to a school holiday or a sick kid at home, and ruthlessly prioritising. Very few of us can claim to be at Julius Ceasar level of multi-tasking. Flexibility of remote work ≠ combining full time work with full time parenting

26

I find Mark's prediction about mid-level software engineers becoming obsolete by 2025, or signals from other companies like Salesforce on total freeze for hiring software engineers, both fascinating and a bit concerning [watch Matthew Berman's commentary - link in the comments]. The biggest challenge I see isn't just job displacement - it's the learning curve. AI is most powerful when you already have the expertise to catch its mistakes. So sure, people who are middle level have already strong foundations to apply their skills elsewhere and redefine their jobs - most likely in amplifying outcomes of the work they've already been doing, now with the use of AI agents. But how do new developers build that foundation if they're immediately relying on AI? It would be like trying to learn a language solely through a translation app. You can fake it only this far. How do we maintain code quality when fewer humans truly understand the systems they're building? The debugging process becomes exponentially more complex. What happens to career progression? If we remove the middle rung of the ladder, how do junior developers gain the expertise needed to become seniors? And how do we ensure AI remains a tool that enhances rather than replaces human understanding and makes us lazy and stupid? For those considering software engineering now, I'd strongly encourage focusing on systems thinking and architectural design - skills that AI still struggles with. Learn to read code critically and understand why certain solutions work (or don't). I think, ultimately the future isn't about competing with AI - it's about mastering the skills AI can't replicate: critical thinking, system design, and problem definition. This has always worked to our advantage across centuries and pivotal moments liek the Industrial Revolution, so I hope it will continue to do so (and the dark scenarios from sci-fi movies about AI taking over won't come true 🤞🏼🫠). PS. I actually explore these themes more deeply in my recent #hankka post "When to use AI, and when to avoid it (and how I use it)" where I share practical strategies for using AI as an amplifier rather than a replacement for human expertise [link in the comments].

15

1 Comment

Anthropic released another paper on their ongoing work towards interpretability of large foundational models like their Claude 3 family, built on previous work by many. Their work continues to be enlightening and inspiring. The findings in this work are generally similar to previous findings on smaller models but scaled up to Claude 3 Sonnet: 💡 they find large numbers of "features" corresponding to combinations of neuron activations and these features map to concepts that we recognize/understand: for example, features for transit infrastructure, particular people, concepts (like bias and "objects lacking sentience"). This might sound familiar from classical machine learning, but the interesting thing is that there is no single thing/neuron encoding each of these concepts, it's a particular combination of neuron activations that embeds these concepts in "superposition" in the network. 💡they observe a sort of "distance" of these feature activations that arrange themselves "near" semantically similar features. When they use their sparse autoencoder approach to learn more and more features they observe that the level of abstraction in the detected features becomes more granular, in ways that are interpretable and meaningful. 💡when they artificially fire these activation patterns during inference, it affects the output in ways that relate to the "feature". For example, asking Claude "what's the most interesting science" normally it responds Physics, but if you clamp the feature activations for the feature "Brain Science" to 10x their max values then Claude responds: Neuroscience. Ways to interpret _and_ influence the model. Great readable work [1] along with an interactive explorer of some of the "features" themselves [2] 🎂 There's an easter egg too: if you scroll in the feature explorer tool and try and select one of the "Randomly selected features from 1M" you get a message "if you want to see the rest of the features, we're hiring!" with a link to their recruiting page :) [1] https://lnkd.in/gZbparwp [2] https://lnkd.in/gC9eus53

19

Recently, Eric Ignasik and Michael Sols from The Software House asked me about my first steps as a VP of Technology at Habito. It's been a fun 8 months in a supportive environment with fantastic people! 😉 This interview discusses a few improvements ‌I made with a big emphasis on why servant leadership is important to me. Hope you’ll learn from it, ➡️ https://hubs.la/Q02Bw-Bv0 #softwaredevelopment #cto #softwaredelivery

67

4 Comments

The new strawberry model from ChatGPT is way more "clever" than previous iterations. The best thing it's doing is normalizing the time spent processing and re-evaluating the steps before spitting out an output, which is vital for correct results to hard problems. While it's addicting to get the real-time stream-of-thought from ChatGPT, getting fewer hallucinations and more accurate answers will be better for everyone... At least until Open Source cracks the next evolution to get the same accuracy at lightning-fast speed!

24

Thank you so much, Jeremy Kahn for the great conversation this week at Fortune #BrainstormAI on the future of Google Search and how we’re expanding what’s possible with AI. One of the most exciting parts of working on Google Search is the immense opportunity we have with AI to be even more helpful – the truth is, there aren’t just billions of questions that people ask Google every day, there are billions of questions that people don’t ask at all – because it’s too hard to formulate the question, or they aren’t sure whether they’ll get a useful answer. In this next era of Search, we really want to  make it truly effortless to ask any question, make sense of any type of information, across all modalities – whether that’s video, image, or text. That includes asking your question in any way that’s most natural to you - think about Google Lens, where you can search what you see, and get an AI Overview with context and links to better understand it. There’s an even greater opportunity as our AI models continue to improve. Just this week, the Google DeepMind team showed off a range of new capabilities with Gemini 2.0 (https://bit.ly/3ZU3JVe), which we’re already testing in AI Overviews so people can tackle even harder questions. And the Android team shared a glimpse into how we’ll help people better understand and interact with the world around with the new XR operating system (https://bit.ly/3ZRx9TR). What’s most exciting to me is how AI is unlocking entirely new ways to make information useful, so regardless of what language you speak, or your skill level in a particular domain, we can help you understand information effortlessly. I can’t wait for the year ahead for Search – there’s so much more to come! You can check out our full conversation at Fortune Brainstorm AI here: https://bit.ly/3DmPdMA 

361

3 Comments

The most clear, fact-based, logically reasoned (and well-written) view on the the positive places that AI can take us from someone who really knows what he's talking about, Dario Amodei (Founder and CEO of Anthropic). This is the article I would have aspired to write if someone asked me to write about this topic. https://lnkd.in/g-xN7zRW

7

2 Comments

Reading Ted Mabrey's article on FDEs at Palantir (https://lnkd.in/dEeweq2g) I'm completely in agreement with his point that you've never "finished" building a product, thus your mission is to expand the "perimeter" of your product as fast as possible. IMO one can see this as a bit like the universe: it's always expanding. Not only do you want to keep expanding the use cases your product addresses, but there are ever more use cases popping up. Going further, if we imagine the number of use cases as a linear dimension, and the effort required as the surface area, unsurprisingly adding more use cases to your already complex product takes much more than linear effort... So not only is a product never finished, it becomes ever harder to expand its perimeter. Thus, building an ecosystem of products (from partners, or a well-integrated solution of one company's products) is likely to work better than one "super app.". The former is a collection of small spheres, each of which can expand to targeted incremental use cases with modest effort, versus the latter which is a huge and very hard to grow.

8

How do you tackle technical debt? Every software company has technical debt, and the majority struggle to pay it back. After reading Shape Up, we've been trialing 8-week cycles of a 6-week build phase followed by a 2-week cool down. The cool down phase allows us to catch our breath and prepare for the next cycle. More importantly, it also allows us to focus on technical debt. The 2-week window gives us plenty of dedicated time to focus on large issues that will have significant benefits for future cycles.

12

5 Comments

The Smartling team couldn't be more excited to introduce the AI Translation Toolkit: a suite of AI translation features designed to reduce translation costs and improve linguist performance. This feature allows #localization teams to leverage AI to streamline translation processes, without sacrificing quality. Reach out to your Smartling Customer Success Manager to learn about how the AI Translation Toolkit can improve the performance of your translators. #aitranslation #smartling #l10n #ailocalization #translatortools

28

It’s eye-opening to see how AI model businesses are dealing with tight margins despite big investments. 🤯 Aidan Gomez’s comments about the “zero margin business” highlight the fierce competition and price cuts in the industry. 💸 With giants like OpenAI and Anthropic lowering prices to stay ahead, it’s a tough time for startups. The future of AI might depend on finding new ways to make these technologies profitable. 🚀 #AI #MachineLearning #TechTrends #Startups https://lnkd.in/ejkDu_wT

8

"Your main job is to make your product win." I think this should apply to any product that your company makes. I also like what he says following this statement about what is NOT your main job. I will summarize here: "Your main job is not to make your manager happy, it is not to make execs happy." Also, note, this main job doesn't only apply to PMs, but also to senior Engineering Managers, Tech Leads, and Designers. #makeyourproductwin #leadership

5

2 Comments

Leading a remote engineering team of 28 has taught me one thing—connection isn’t optional. It’s not just about keeping operations running; it’s about intentionally creating moments of connection and trust. As this article points out, remote leadership goes beyond virtual meetings and structured check-ins. Remote work doesn’t mean distant work—it’s just different. #RemoteWork #EngineeringLeadership #RemoteTeamManagement #Startups #WorkLifeBalance #TeamVibes

25

Clear and rich information on RAG variations. I think this will be useful to most practitioners who have gone beyond the intro and demo workflows, and now need to consider how to refine their flows.

10