Tom Ossa

Human being. Owns digital advertising firm & news company. Avid 9w road biker & musician. Enjoys playing chess and cooking. Tries to tell joke, fails. Repeats.

[CA Vivek Khatri @CaVivekkhatri]

Nobody talks about this. A 1998 masterclass where Warren Buffett explains how to never lose money investing.

Watch this before your next trade.

Rare lecture. 1 hour. Pure investing principles. Must Watch.

The game isn’t about making money.
It’s about not losing it when others do.

[Tom Ossa @rocklandweb]

FUCKING EXACTLY.

[CSPAN @cspan]

President Trump on Israel and Iran: "We basically have two countries that have been fighting so long and so hard that they don’t know what the fuck they’re doing."

[Tom Ossa @rocklandweb]

@BusinessWire It is! Effective communication is absolutely key!

Hey, on a side note - Too bad your team lost another client, after not being able to fix that communications feed to them, FOR THREE MONTHS.

But, that’s ok. I see your stats. Probably best they left now.

Take good care. 😁

[Tom Ossa @rocklandweb]

@jsamrad I don’t know, man - even when he’s not so good, he’s pretty good. Check out his ‘19 stats, same ERA as of right now

[SNY @SNYtv]

Edwin Díaz explains why he exited the game in the 10th inning, chalking it up to cramps in his hip

He is not concerned at the moment about this being a problem moving forward

[SNY @SNYtv]

Mets trainers noticed that one of Edwin Díaz's legs was longer than the other yesterday

They fixed the issue, says Díaz: "I was feeling better after"

[SleeperMets @SleeperMets]

Edwin Diaz is exiting today’s game with an apparent injury.

Diaz seemingly pointed to his left hip, directly after disengaging three different times, which led to a balk.

Hoping it’s nothing serious 🙏

[Tom Ossa @rocklandweb]

@WagtheDogPromo @JimNorton Um, usually bees don’t fly into one bedroom basement studios.

[The Babylon Bee @TheBabylonBee]

Local Man Wishes Atheist Would Shut Up About God For Just One Minute https://t.co/Gib4Uwjo00

[ToonHive @ToonHive]

The entirety of ‘Looney Tunes’ has been removed from Max.

[doktor 🪶 @doktor_oa]

“guys let’s release a looney tunes movie and then remove any chance of audiences engaging with the rest of the franchise”

wb suits:

[Tom Ossa @rocklandweb]

All the people to NOT vote for in the next election cycle. How disrespectful. And you call us Republicans supposedly rude and crude? We’ e never pulled crap like this at a Congressional address in modern times.

Shame on each and every one of you hypocrites.

[Tom Ossa @rocklandweb]

.@brianglenntv as a Republican - I say this.

You are an absolute dipshit.

[Tom Ossa @rocklandweb]

.@BenBurgis I wouldn’t exactly call it “scared”. More like “case closed” for a while. Since he owns the publication, he does have that right.

That said, the article was very informative and gave us an idea of the news landscape right now.

https://t.co/zsJjciakJD

[Tom Ossa @rocklandweb]

@brianglenntv you’re a fucking dipshit.

[Tom Ossa @rocklandweb]

While science may be empirical, bureaucracy is not.

~ Wags, Billions S7 E4

[Tom Ossa @rocklandweb]

100%. I’ve been waiting for this to finally happen for literally 8 years!

[Elon Musk @elonmusk]

Bruh, if I wanted to rummage through random personal shit, I could have done that at PAYPAL. Hello???

Having tens of millions of people marked in Social Security as “ALIVE” when they are definitely dead is a HUGE problem.

Obviously.

Some of these people would have been alive before America existed as a country. Think about that for a second …

[Jim Fan @DrJimFan]

We are living in a timeline where a non-US company is keeping the original mission of OpenAI alive - truly open, frontier research that empowers all. It makes no sense. The most entertaining outcome is the most likely.

DeepSeek-R1 not only open-sources a barrage of models but also spills all the training secrets. They are perhaps the first OSS project that shows major, sustained growth of an RL flywheel.

Impact can be done by "ASI achieved internally" or mythical names like "Project Strawberry".
Impact can also be done by simply dumping the raw algorithms and matplotlib learning curves.

I'm reading the paper:

> Purely driven by RL, no SFT at all ("cold start"). Reminiscent of AlphaZero - master Go, Shogi, and Chess from scratch, without imitating human grandmaster moves first. This is the most significant takeaway from the paper.
> Use groundtruth rewards computed by hardcoded rules. Avoid any learned reward models that RL can easily hack against.
> Thinking time of the model steadily increases as training proceeds - this is not pre-programmed, but an emergent property!
> Emergence of self-reflection and exploration behaviors.
> GRPO instead of PPO: it removes the critic net from PPO and uses the average reward of multiple samples instead. Simple method to reduce memory use. Note that GRPO was also invented by DeepSeek in Feb 2024 ... what a cracked team.

[Tom Ossa @rocklandweb]

Now, this is something that I can get excited about.

Musk and Ramaswamy Unveil Plan to Overhaul Government

[DHH @dhh]

Maybe I'm supposed to get mad at this, but instead I just get sad. Mullenweg clearly sees it as a failure to create much more value in the world than what you capture, and maybe that's the root of our differences. I see that as a proud achievement. https://t.co/SBDgBE5X10