Martin Lynch

Founder & CEO of XLR8·4ward (bringing usable AI to insurance). Interested in the intersections of finance & tech, AI & humanity, & quantum physics & philosophy.

[Martin Lynch @MartinMLynch]

@vawkeicodewebz @garrytan I'm the founder of XLR8•4ward.

We make an AI-powered underwriting platform (InfiniteUnderwriter) for the P&C insurance industry, which I spent a few decades in.

The industry is powered on profoundly obsolete tech.

Every day of my career I've wanted to fix that.

Now I am.

[Martin Lynch @MartinMLynch]

@CXCarroll Tangentially related: many people assume McDonald's business is selling burgers & fries. It isn't.

They are a giant real estate company.

The bulk of their profits are earned from the rent they charge ~22,000 franchisees for their buildings and land, which McDonald's owns.

[Martin Lynch @MartinMLynch]

@PeterDiamandis Given how profoundly inefficient humans are at learning (it takes us *years* to learn to use a bathroom properly) almost anything we can do to outsource intelligence and knowledge has massive returns compared to the "manual labor" that is university learning.

[Martin Lynch @MartinMLynch]

@paulg I've wondered the same about the few people I used to encounter in my Cybertruck who would give me the finger, which never happens anymore.

What caused them to decide to stop giving the finger to strangers based on a vehicle? Do they ever reflect on how odd that is?

[Martin Lynch @MartinMLynch]

@Rainmaker1973 It remains the case that the entire population of the world could fit into Texas with each person having about 1,000 square feet (~7.5 trillion square feet divided by ~8 billion people).

The world is big. It just doesn't seem that way.

[Martin Lynch @MartinMLynch]

This is precisely what my company's AI tools do: they remove from our customers the tedious work (like data entry & information collection) that consumes time away from the thinking they are paid to do.

Prior attempts to abstract away this work by giving it to less skilled support staff doesn't really work because the time spent communicating those tasks often exceeds the time to do it themselves, and introduces errors.

[Martin Lynch @MartinMLynch]

@PeterDiamandis This is a vital point: all progress comes from going out to where you don't have the answers, but having the conviction that you'll be able to solve the problems as they arise.

Its polar opposite is bureaucracy: attempting to determine all the answers before the journey starts.

[Martin Lynch @MartinMLynch]

@gilbert My brain has a neighboring place for, "You've copied something to the clipboard - hurry up and paste it before it disappears!"

[Martin Lynch @MartinMLynch]

@GaryMarcus Probably true.

Except at the extreme edges, AI models are mostly identical for everyday purposes. They are thus far unprofitable & subject to massive capex in a commoditized space where any margins will trend towards zero.

I don't think the 1st to self-improvement takes all.

[Martin Lynch @MartinMLynch]

@paulg The attempt to compress an idea into 280 characters here helps hone clarity, and its cousin, brevity (this post was originally 18 muddled paragraphs...).

[Martin Lynch @MartinMLynch]

LeCun is right: for us, language is a lossy compression mechanism of reality. But for LLMs, it *is* their base reality.

And language is *very* lossy: you can use a million words to describe a particular rock, but only the rock itself contains its full description.

That doesn't mean LLMs cannot be extremely useful, it just describes their limitations.

Multimodal training using inputs equivalent to all our senses probably gets you there, but that's an exponentially greater task than using the shortcut of words.

[Martin Lynch @MartinMLynch]

@LoganDobson The AI revolution will almost certainly be a massive net benefit to most, but it's worth remembering the 'Engels pause':

It took about 50 years for the industrial revolution's effects to be net positive for most.

That delay will certainly be much shorter with AI, but not zero.

[Martin Lynch @MartinMLynch]

@heynavtoor This is not at all surprising.

Our brains don't take as their input sights, sounds, touch. They don't "see" the Pong game either.

Their only input is electrical impulses. They decide some are signal, others, noise.

From that we build our worlds. And play Pong.

[Martin Lynch @MartinMLynch]

I asked Gemini to do a post mortem on how I could have provided better instructions. Its response:

"Your instructions were clear, but they allowed me too much interpretative freedom. You must treat AI as a Junior Coder who is prone to lying."

Noted.

[Martin Lynch @MartinMLynch]

It's hard to overstate just how bad Gemini 3 Flash Preview is.

Spent the day coding with it. It's maddening:

- routinely changes variable names crashing downstream code.

- quickly loses track of what we're doing.

- routinely ignores instructions.

I sure miss Gemini 2.5 Pro.

[Martin Lynch @MartinMLynch]

@Jason We make an AI-powered P&C underwriting platform.

The resultant opportunities opening up are at the top of the funnel: business development & leads, because our platform has solved the bottleneck at the bottom of the funnel.

Filling the top at speed is now even more valuable.

[Martin Lynch @MartinMLynch]

One of the best features of Chrome is the ability to bookmark *all* your open tabs in a single bookmark (option in the "Bookmark" menu).

I usually have 4+ windows with 25+ tabs each so I'll bookmark all tabs in each window.

To open, right click the bookmark and select "open in new window".

Eliminates the anxiety Paul describes (which I used to share).

[Martin Lynch @MartinMLynch]

We're re-living the early reaction to the industrial revolution: the rise of the Luddites, fears that new tech will displace all workers, attempts to legislate away impending progress.

That transition was genuinely painful for some.

This one will be too.

But net positive for most.

[Martin Lynch @MartinMLynch]

@unusual_whales Dimon isn't stupid or obtuse, he's defending his realm.

Banks at root sell trust.

Bitcoin and cryptos obviate the need for it.

They are an existential threat to bankers, coming at them in slow motion.

His efforts to delegitimize make sense (but at best will stall things).

[Martin Lynch @MartinMLynch]

When we look back 100 years at doctors using bloodletting, injecting malaria to supposedly cure syphilis, performing lobotomies to treat depression & think it all absurd our immediate reaction should be to wonder which things we do now will be thought absurd 100 years from now.

Everything feels cutting edge in the moment. Advances prove some of it wrong. There's a return in figuring out the wrong parts faster.

[Martin Lynch @MartinMLynch]

@Scobleizer It varies.

We've worked with large companies ready to go all-in on an AI pivot.

But also those with very deep institutional resistance to change, often well below the executive level (the execs want AI to work), where the appearance of engagement masks flying monkey wenches.