frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Palantir's secret weapon isn't AI – it's Ontology. An open-source deep dive

https://github.com/Leading-AI-IO/palantir-ontology-strategy
65•leading-AI•2h ago

Comments

jey•1h ago
This seems to be the English landing page: https://github.com/Leading-AI-IO/palantir-ontology-strategy/...
est•1h ago
It's just view, materialized view, udf, stored procedure in fancy corp speak.
trollbridge•1h ago
Believe it or not, this stuff is still incredibly valuable.
est•1h ago
I have to admit, those fancy concepts pay the bills.
trollbridge•22m ago
I'm having a blast with using PostgreSQL to hold unstructured data in JSON-B and then creating views off of that that "feel" like more conventional SQL.

If performance becomes an issue, just turn it into an MV... and then consider some indexing on the JSON itself.

kumarvvr•1h ago
Do you have any resources or books to learn all the details of all these?

Also, what is UDF?

tfryman•1h ago
User defined functions
AxiomaticSpace•1h ago
Pretty much any SQL book will cover it those, and there's a bunch of online SQL tutorials. UDF means user defined function, so if there's some function you want to perform in SQL but that function doesn't come out of the box, you can just write your own. And those can be defined in non-SQL syntax, such as UDF's written in python or C++, which can be pretty handy.
dmoy•1h ago
> python or C++

Also sometimes Lua, which is kinda a nice middleground between c++ efficiency and python ease of writing

whattheheckheck•51m ago
Data warehouse toolkit for Kimball

Google will inmon for the info factory

Google data vault

Joe reis is the guy for tying it up with a modern bow recently.

Designing data intensive applications book

rapind•16m ago
> fancy Karp speak.

The paypal mafia are all about stories. They can attract talent and investor money with those stories, but they are just tall tales full of hype, and people are catching on (ok that last bit might be a hopium).

DauntingPear7•1h ago
Reads like AI
locusofself•1h ago
I wonder if Michael Bury's puts printed. The stock is down about 30% from when I think he announced them.
gaigalas•1h ago
> 2-1. Modeling the Data World with "Nouns" and "Verbs"

> Link type: The relationships between object types, supporting 1-to-1, 1-to-many, and many-to-many relationships.

Seems incredibly naive in terms of symbolic representation of knowledge. Maybe I spent too much time with OWL.

CGMthrowaway•1h ago
Can you say more? What is state of the art?
gaigalas•1h ago
I've learned this stuff as a hobby, so take it with a grain of salt. I'm not a specialist.

OWL 1, for example, has stuff like transitive properties (the classical example is A ancestorOf B, B ancestorOf C, therefore I can infer A ancestorOf C if I annotate ancestorOf as a transitive property).

Union, equivalence, inversion, symmetries, cardinality. Those are all possible to represent symbolic in OWL ontologies.

They're also neatly separated in different types (OWL Lite, OWL DL, OWL Full). OWL Lite and DL for example are proven to be decidable (you won't get some halt when doing inference, no matter what).

I know there are plenty of database engines to store triples and graphs, and plenty of reasoners out there.

I haven't studied OWL 2 yet or newer stuff like SHACL, but I know it's supposed to be even better.

HillRat•1h ago
I find myself distinctly unimpressed by the idea that slapping a nice UI and some TS/SCI controls on top of a graph database — the latter being something that NSA did, with considerably more sophistication, years prior in a Neo4J fork — is some kind of brilliant conceptual moat. Graph DBs are useful for certain kinds of problems, which happen to map well to counterterror social mapping strategies, this is nothing particularly new or noteworthy.
measurablefunc•1h ago
And my secret is epistemology. AMA.
WaitWaitWha•1h ago
mine is axiology, DNAMA. ;)
bryanrasmussen•1h ago
well my MA said my DNA is secret.
notepad0x90•1h ago
palantir doesn't do revolutionary things in terms of back-ends. matter of fact, their apps are at best mid. I'd rate them 3/10 compared to alternatives that can do similar things. Their front end is the real differentiator.

Their bread-and-butter is a few things

1) Willing to do dirty/harmful things no one else will touch

2) Making data and data analysis accessible to cops, dhs, anyone that is especially tech-averse (many police departments disqualify based on IQ test results measuring too high). You can type in a license plate, a name, an address, scan a face and it will show you every relevant information, but also contextualizes it and enriches it with any other data. You could to this in excel, postgresql, bigquery, etc.. but palantir gives these people simple text boxes, buttons, and links.

3) Their forward deployed engineers are great at what they do. They station their guys wherever Palantir is being used, and they'll work very closely to get things done. to make sure all problems are solved asap, and its users are very well educated on the usage of the platform.

This post looks like it's written by AI, but assuming it is in earnest, it isn't really ontology, at least no more than object oriented programming is ontology. Excel is all about numbers, palantir is all about people (or people-documents). It is simpler than excel and has BigQuery level analytical power behind it, and the human touch to make that interaction go over really well.

I said it's mid because you could do a lot more with just the dataset and queries. You could even possibly do more with command line tools and hoards of data files (minus the OCR and document scanning they do, as well as LLM/NLP). but that isn't accessible and takes a lot more time. Not to mention normalizing, extracting and structuring wildy unstructured data isn't easy. But with BigQ for example, it is done plenty, you just hire a team to do that for you typically.

Their ecosystem is basically google search (including image, reverse image,video,etc..) but much more targeted and oriented towards displaying collated data from hoards of structured and unstructured data (including pdfs, docx,etc..). I would prefer grep, bigquery,splunk myself. but for end users, palantir is unmatched in my experience.

But I'm not selling them here, I'm trying to communicate the power at the disposal of those who use palantir's platforms. Google could have crushed them any time, except even for Google the type of work required was too ghoulish and reputationally risky.

Even with MS copilot(lol), chatgpt, gemini,etc.. running as agents, they're not as simply as palantir's stuff is for searching your data. and you don't have specialists integrating all your data onsite either.

Ultimately, the bigger problem is that even in crowds like HN's, no one seems to have a good idea of what should be done about governments abusing datascience so efficiently. Every answer comes back to red-tapes and regulations, possibly criminal consequence. Are you willing to give up the liberties tech has enjoyed so that future generations can be well, and have shot at peace and prosperity? (ours is too far gone in my opinion)?

China is doing this too, but much more efficiently, much better and at a greater scale. but their society has accepted this, and traded certain liberties for social stability and economic prosperity. The west hasn't done that. lawmakers and the public at large need to be informed by those in tech about these things so informed decisions could be made.

qlfhvt76b•1h ago
It doesnt matter who does search well, if you understand the implications of the theory of bounded rationality and what claude shannons information theory tells us about bandwith limitations of individual and group cognition.
rramadass•58m ago
> Ultimately, the bigger problem is that even in crowds like HN's, no one seems to have a good idea of what should be done about governments abusing datascience so efficiently.

The general-populace/crowd/mob has already lost this game. Govts/Companies (all of them irrespective of ethics/democracy/etc.) are doing what they want with data and datascience. The populace is easily propagandized/distracted from reality and can be easily cocooned.

The only recourse left for the individual is to learn and start playing the game himself. Fortunately the new tools are a great help in this asymmetric warfare. Organizations like EFF/OpenSource/GNU/etc. need to take the lead on this since most people are like sheep when it comes to uses/misuses of technology.

Palantir is just "Cambridge Analytica" redux but with more money/connections/data/breadth/depth/etc. Watch this old presentation by their then ceo Alexander Nix and extrapolate to today's AI world - https://www.youtube.com/watch?v=n8Dd5aVXLCc

Finally go and read the works of George Orwell, Edward Bernays, Jacques Ellul, Marshall Mcluhan, Noam Chomsky etc. on the whole subject of Propaganda/Manipulation to really understand where we are now.

bluegatty•1h ago
This feels like pets.com made by Accenture consultants.
echelon•1h ago
Palantir's secret weapon is the closeness and affinity to the DoD.

The tech stack ontological model is flexible like Salesforce so that it can be jammed into any task or contract quicky. It isn't engineered, it's glued in.

They're able to do this fast because they have a flexible model and because they have the friendly relationships.

Their moat deepens every year with every new integration.

It's smart as hell, actually. That's why they're swimming in money. And government contracts are about as lucrative as you can get.

Engineers turn their nose at this, but look who has tapped into this wealthy revenue stream. While we preen about good architecture, they can retire for a thousand thousand lifetimes.

bluegatty•57m ago
Palantir does not have infinity money, and 'ontology' is a buzzword.

They're a gov. contracting agency, with some re-usable components, that's it.

If they deliver stuff that works, good, if not, bad.

There's nothing interesting about 'ontology'

wbsun•31m ago
> Palantir's secret weapon is the closeness and affinity to the DoD.

Is it a secret? I got an impression that it has been well known. How could you get any big number contracts without former secretaries or retired generals in your board or in your ‘consulting’ team?

runarberg•30m ago
> Engineers turn their nose at this, but look who has tapped into this wealthy revenue stream.

This may be one of the most tone deaf, american imperialist sentiment, I’ve heard on HN for a while.

Engineers who have any sense of morality have a pretty good reason to turn their nose at this, and there is no but needed to follow that sentence.

zdragnar•14m ago
If you read the comment a little more closely, it is very obvious that the "this" engineers turn their noses up at is the flexible model full of glue code, ala Salesforce, as opposed to "good architecture".

It's more or less in the same vein as pointing out that WordPress powered a massive chunk of the Internet despite violating almost every good coding practice you can name, and that getting things done is what makes money, not building ivory towers.

The fact that you turned that argument into some sort of anti American screed says much more about you than the parent.

runarberg•4m ago
That is why I called it tone deaf, I admit the part about American Imperialism may have been unwarranted (may is in emphasis for a reason).

This engineer turned their nose at the bad architecture and glue code, but neglected to mention the total lack of morality from Palantir. I would argue that abandoning morality and aiding the American imperalist machine in its war against human rights and dignity, has been a much bigger reason for Palantir’s success then their lack of good engineering practice. They are willing to get paid for something most people morally object to. Lots of engineers are willing to abandon their craftsmanship if it pays well enough, few their morals.

Perhaps I read too much into this absence, in which case the post is only tone deaf, but I favor the read where this absence was intentional, in which case it is both tone deaf and American imperialist.

m00dy•1h ago
is it accurate ?
WaitWaitWha•54m ago
I really wanted this "book" to be good.

In the context of the paper, the entire book seems to go downhill from the definition of ontology for me.

There is no benefit of using Gruber's ivory tower definition. A simpler explanation (e.g., it describes a structured framework that defines and categorizes the entities within a specific domain and the relationships among those entities) would have sufficed, and easier to digest.

Palantir is doing nothing revolutionary or "paradigm shift" when it comes to data and information organization. Their secret weapon is not introducing ontology to information.

Ching (1000BC?) classified reality into binary ontological primitives, created trigrams and hexagrams a combinatorial ontology. Aristotle introduced categories, substance, properties, relations, etc. Thomas Aquinas systemized Aristotelian categories into theological knowledge systems, and used structured classifications.

I am becoming curmudgeony as I see more and more of these reverse-research papers. Write the paper, then find references that fit the statement and use weasel words ...

unbelievable scene unfolds, deep-rooted disease of silos, paradigm shift, fatal flaws, forged in these extreme environments, eliminated to the absolute limit...

Gag me.

SilverElfin•53m ago
It’s just a consulting firm with connections to the Trump administration through Thiel. That’s why they need forward engineers. It’s not a real platform. Ontology is as far as I can tell, a marketing buzzword, mostly repeated by bots to pump Palantir.
rramadass•53m ago
Palantir is just "Cambridge Analytica" redux but with more money/connections/data/breadth/depth/etc. Their ethical/moral stance i will leave it to you to infer.

Watch these old presentations by CA's then ceo Alexander Nix and extrapolate to today's AI world - https://www.youtube.com/watch?v=n8Dd5aVXLCc and https://www.youtube.com/watch?v=6bG5ps5KdDo (note the Q/A at the end here)

Also watch this interview with Christopher Wylie the CA whistleblower and again extrapolate to today's AI world - https://www.youtube.com/watch?v=FXdYSQ6nu-M

Be afraid, Be very afraid.

umairnadeem123•28m ago
The real insight buried in this thread is that Palantir's moat has nothing to do with ontology as a concept - it's that they solved the data integration problem for organizations that are deeply allergic to engineering. Every large org I've worked with has the same issue: dozens of data sources, no consistent schema, and teams that refuse to standardize. Palantir just charges a premium to do the dirty work of normalization and relationship mapping that most internal teams can't get funding or political buy-in for. You could replicate the tech with a graph DB and some ETL pipelines, but good luck getting a government agency to build and maintain that internally.
mannyv•27m ago
I worked at a company where the VPs etc spewed the word "ontology" all the time. We were never sure if they knew they were spewing bullshit or they really believed what they were saying was real.

Ontology is one of those fancy words that sounds important but is basically, as another poster pointed out, a standardized vocabulary.

modeless•22m ago
> isn't X – it's Y

What is this AI slop doing at the top of HN? Come on, you don't even have to click through to know it's slop! It even has an en dash right in the title!

pcurve•16m ago
I think their secret weapon is their opensource UI library.

https://blueprintjs.com/

I've been tracking this project since 2018. It hasn't changed drastically since then, but man it was polished and robust library.

gigatexal•12m ago
Eccentric CEO (we already have Elon another is one too many …) wildly overpriced stock and company and also one without any ethics that’s super close to the Trump admin … I for one am praying this company collapses. I’d love to read a HN post about how they go chapter 11 and leave the public eye.

How I use Claude Code: Separation of planning and execution

https://boristane.com/blog/how-i-use-claude-code/
238•vinhnx•4h ago•144 comments

Japanese Woodblock Print Search

https://ukiyo-e.org/
25•curmudgeon22•1h ago•6 comments

Show HN: Llama 3.1 70B on a single RTX 3090 via NVMe-to-GPU bypassing the CPU

https://github.com/xaskasdf/ntransformer
152•xaskasdf•8h ago•34 comments

A Botnet Accidentally Destroyed I2P

https://www.sambent.com/a-botnet-accidentally-destroyed-i2p-the-full-story/
52•Cider9986•3h ago•21 comments

Evidence of the bouba-kiki effect in naïve baby chicks

https://www.science.org/doi/10.1126/science.adq7188
98•suddenlybananas•7h ago•25 comments

How far back in time can you understand English?

https://www.deadlanguagesociety.com/p/how-far-back-in-time-understand-english
428•spzb•3d ago•242 comments

Two Bits Are Better Than One: making bloom filters 2x more accurate

https://floedb.ai/blog/two-bits-are-better-than-one-making-bloom-filters-2x-more-accurate
34•matheusalmeida•4d ago•3 comments

Scientists discover recent tectonic activity on the moon

https://phys.org/news/2026-02-scientists-tectonic-moon.html
27•bookmtn•4d ago•2 comments

“Playmakers,” reviewed: The race to give every child a toy

https://www.newyorker.com/magazine/2026/02/16/playmakers-the-jewish-entrepreneurs-who-created-the...
4•fortran77•1d ago•0 comments

zclaw: personal AI assistant in under 888 KB, running on an ESP32

https://github.com/tnm/zclaw
133•tosh•16h ago•73 comments

Parse, Don't Validate and Type-Driven Design in Rust

https://www.harudagondi.space/blog/parse-dont-validate-and-type-driven-design-in-rust/
148•todsacerdoti•9h ago•39 comments

Gamedate – A site to revive dead multiplayer games

https://gamedate.org/
15•msuniverse2026•1d ago•1 comments

People Loved the Dot-Com Boom. The A.I. Boom, Not So Much

https://www.nytimes.com/2026/02/21/technology/ai-boom-backlash.html
40•1vuio0pswjnm7•1h ago•32 comments

Forward propagation of errors through time

https://nicolaszucchet.github.io/Forward-propagation-errors-through-time/
18•iNic•2d ago•0 comments

CXMT has been offering DDR4 chips at about half the prevailing market rate

https://www.koreaherald.com/article/10679206
176•phront•14h ago•157 comments

Claws are now a new layer on top of LLM agents

https://twitter.com/karpathy/status/2024987174077432126
236•Cyphase•1d ago•673 comments

Toyota Mirai hydrogen car depreciation: 65% value loss in a year

https://carbuzz.com/toyota-mirai-massive-depreciation-one-year/
120•iancmceachern•10h ago•274 comments

EDuke32 – Duke Nukem 3D (Open-Source)

https://www.eduke32.com/
164•reconnecting•8h ago•61 comments

Canvas_ity: A tiny, single-header <canvas>-like 2D rasterizer for C++

https://github.com/a-e-k/canvas_ity
73•PaulHoule•10h ago•25 comments

I verified my LinkedIn identity. Here's what I handed over

https://thelocalstack.eu/posts/linkedin-identity-verification-privacy/
1202•ColinWright•21h ago•424 comments

Inputlag.science – Repository of knowledge about input lag in gaming

https://inputlag.science
75•akyuu•9h ago•13 comments

Be wary of Bluesky

https://kevinak.se/blog/be-wary-of-bluesky
274•kevinak•1d ago•181 comments

What not to write on your security clearance form (1988)

https://milk.com/wall-o-shame/security_clearance.html
413•wizardforhire•11h ago•183 comments

Finding forall-exists Hyperbugs using Symbolic Execution

https://dl.acm.org/doi/full/10.1145/3689761
24•todsacerdoti•5d ago•1 comments

Uncovering insiders and alpha on Polymarket with AI

https://twitter.com/peterjliu/status/2024901585806225723
140•somerandomness•1d ago•130 comments

A16z partner says that the theory that we’ll vibe code everything is wrong

https://www.aol.com/articles/a16z-partner-says-theory-well-050150534.html
103•paulpauper•1d ago•165 comments

Personal Statement of a CIA Analyst

https://antipolygraph.org/statements/statement-038.shtml
198•grubbs•11h ago•120 comments

Permacomputing

https://wiki.xxiivv.com/site/permacomputing.html
117•tosh•4d ago•28 comments

Keep Android Open

https://f-droid.org/2026/02/20/twif.html
2017•LorenDB•1d ago•694 comments

The importance of limiting syndication feed requests in some way

https://utcc.utoronto.ca/~cks/space/blog/web/FeedLimitingImportance
3•LorenDB•9m ago•0 comments