Instagram story viewer> @systemsluke> Posts

Lukasz Ksiazczyk

@systemsluke

559

followers

156

following

building systems
ops manager by day
talking about ai, tech and systems

POSTS STORIES REELS TAGGED

Download All

@c.h.robinson, a logistics company that managed to grow its operating profit while the wider freight industry was struggling.

They use a framework called Lean AI. The basic idea is to understand exactly how work moves through the business before deciding where AI should be involved.

That means finding the quotes sitting unanswered, the information being entered twice, the unnecessary handoffs and all the small delays that quietly make the business slower. Once those problems are clear, they can build AI agents for specific jobs, such as pricing shipments, processing orders or checking missed pickups.

Since the end of 2022, productivity across the company’s two main divisions has increased by more than 60%.

It’s a useful reminder that choosing the right problem matters just as much as choosing the right technology.

#AI #logistics #automation by @systemsluke

4

0

a day ago

Technology normally gets cheaper over time. But AI is moving so quickly that the industry is already spending billions on the next breakthrough before the current one becomes affordable.

That race needs more chips, memory, data centres and electricity. We’re already seeing the effects through rising RAM prices, expensive graphics cards and AI subscriptions that can cost hundreds every month.

There is still another possible outcome. More factories, better efficiency, smaller models and stronger competition could bring these costs down.

This video looks at both sides and asks what happens if the frontier of technology keeps moving faster than most people can afford to follow it.

#Tech #AI #RAMPrices by @systemsluke

5

0

2 days ago

Companies are using AI to increase productivity and reduce headcount. But if two competitors use AI equally well, surely the company with more AI-equipped employees still has more capacity to build products, test ideas and serve customers.

Early research supports this. A study at Procter & Gamble found that one person using AI could perform about as well as a two-person team without it. However, teams using AI produced the strongest results.

So perhaps today’s layoffs are an adjustment period. Once companies have reorganised around AI, the advantage may belong to those combining the technology with more talented people, rather than simply cutting the most jobs.

#layoffs #AI #tech by @systemsluke

3

1

3 days ago

Meta’s latest releases feel like the clearest sign yet that the company is building beyond the smartphone.

Its new VR glasses look closer to what many people expected from the Apple Vision Pro, while Muse Charm brings AI into a more playful, personal product, almost like a modern Tamagotchi.

What’s interesting is how everything is starting to connect. Meta’s AI, glasses, wearables and social platforms no longer feel like random experiments. They’re beginning to look like one bigger vision for how we might communicate, create and interact with technology after smartphones.

#Meta #VR #MuseCharm by @systemsluke

1

0

4 days ago

I’ve been looking into how new AI models are actually tested, because benchmark scores are usually one of the first things people check when a new model is released.

Benchmarks give different models the same tasks and score them using the same rules. This helps us compare their abilities in areas like coding, maths, and scientific reasoning.

The problem is that models are improving so quickly that some tests are struggling to keep up. Researchers recently reviewed several advanced physics benchmarks and found unclear questions, incorrect reference answers, and grading mistakes. Once corrected, the results showed that the models were far more capable than the original scores suggested.

This is creating opportunities for companies like Snorkel, which builds training data, practice environments, and better ways to measure performance.

Sometimes the biggest opportunities aren’t in the new technology itself. They’re in the tools that help us train it, test it, and understand what it can actually do.

#AI #Benchmarking #Testing by @systemsluke

9

0

5 days ago

Revolut is a great example of how you don’t need to reinvent an entire industry to disrupt it. They found one problem people already hated, solved it better than traditional banks and used that as their way into the market. Now, roughly one in four UK adults uses Revolut, and the banking giants are being forced to rethink how they compete.

#banking #Revolut #growth by @systemsluke

3

0

6 days ago

Spending time online can make it feel like everyone is already using AI to automate their work, build businesses and completely change the way they live. I wanted to find out how true that actually was, so I looked through some of the latest research on AI adoption.

The numbers tell a very different story. Most people still aren’t using AI regularly, and plenty of people have chosen not to use it because they aren’t interested or they have concerns about privacy and accuracy. At the same time, the people who use it heavily are starting to hand over much larger and more complicated tasks.

So if you feel behind, you probably have more time than you think. You don’t need to understand every new model or try every tool. Start with one task you already do regularly, see whether AI can help, and judge it by the result.

#AI by @systemsluke

3

0

7 days ago

The more I looked into Jev, the more I understood why people are paying attention to it.

Most of the AI tools we use today were built to create language. They read what you send them, then build a response piece by piece until they have a complete answer. That works well when you want to chat, write something or work through a complicated problem. But sometimes you only need AI to make a quick choice, and a full written response adds time and cost that the software doesn’t really need.

That’s the idea behind Jev, created by TypeSafe AI after its founder left OpenAI. Instead of writing a long answer, Jev is given a set of possible choices. It scores them, selects one and tells you how confident it is. It can also handle several decisions at once.

TypeSafe says this approach can be up to 200 times faster and over 400 times cheaper for the kinds of tasks Jev was built to handle. It won’t replace ChatGPT because it isn’t trying to do the same job. The interesting question is whether this becomes another layer of AI that works quietly inside the apps and services we already use, making thousands of small decisions without turning every one of them into a conversation.

#AI #TypeSafe #OpenAI by @systemsluke

7

0

8 days ago

Have you heard of Bending Spoons? Most people haven’t, but they quietly own Evernote, WeTransfer, Vimeo, AOL and Eventbrite. Now they’ve bought Airtable and agreed to acquire Miro too.

What’s fascinating is that they’re not trying to connect all these products into one big software suite. Their real strategy is to buy well-known companies with predictable revenue, rebuild how they operate, improve the margins and use the profits to fund the next acquisition.

It’s an aggressive model, and it often involves major restructuring, but it’s turning Bending Spoons into one of the most interesting companies in technology. They’re essentially building a machine for acquiring and transforming software businesses, and their ambitions seem to be getting much bigger.

#BendingSpoons #Tech #Software by @systemsluke

19

2

9 days ago

OpenAI just launched Astra for Law, and I think it’s a smart way into the legal industry.

It’s basically GPT-6 Astra tuned for legal work. It searches an index of over 230 million URLs, so lawyers can research cases, analyse documents and draft arguments.

The interesting bit is Harvey and Legora. They already sell AI tools to law firms, and now they can build Astra for Law into their own products through the API.

So OpenAI ends up as the intelligence running underneath the platforms lawyers already work in. ChatGPT, Codex, in-house tools, specialist platforms: that’s a lot of routes into one industry.

#AI #OpenAI #AstraLaw by @systemsluke

3

0

10 days ago

8

0

11 days ago

So much of the AI conversation is focused on tools for businesses, from automating workflows to reducing costs and improving productivity. But Brian Chesky, the CEO of Airbnb, recently raised an interesting point. The bigger long-term opportunity could be consumer AI and the ways it becomes part of everyday life.

That might mean conversations with AI becoming more visual and interactive, wearables providing useful information in real time, or personal agents quietly helping with tasks in the background. The real shift happens when AI stops feeling like a separate tool we have to open and starts fitting naturally into how we live, make decisions and interact with the world.

#AI #Airbnb #Tech by @systemsluke

3

0

12 days ago