The jam of the turnip; I do SEng stuff. Mostly Android, iOS, and Spring. Sometimes rt silly stuff
A Mason Jar
Joined July 2013
- Tweets14.9K
- Following101
- Followers304
- Likes176K
Bro the ending
My dad got me taken by Canadian border patrol when I was 5 because they thought he was smuggling chocolate into the country.
My parents, my siblings and I, were passing through Canada to reach Alaska. At the border, the patrol were asking dad typical questions for someone entering. Any firearms? No. Fireworks? No. Chocolate? My dad just yells out CHOCOLATE?! REALLY?! Well… maybe we got some chocolate cake left in the fridge but CHOCOLATE?!.
Patrol surrounded our camper and had dogs sniff around the car. They found the cake, several bars of chocolate, and unmarked beef.
We all got out of the camper and patrol picked me up and walked off with me.
They started questioning me separate from my family in a room. I remember the ceiling was white and the bottom half of the walls was brown.
They asked me who the people in the car were. I told them it was my family. That my dad was the one driving.
Patrol asked if they had kidnapped me and to not be afraid they would take me back to my real family.
I told them again that he was my dad and that was my family.
During this they are questioning my family in a different room asking for ID for my older siblings. My sister had a school id and my brother pulled out a picture of goofy.
Patrol got annoyed and realized they weren’t getting far with us. We were all just a dumb family wasting their time.
They took all our chocolate and meat in the camper then let us go. They insisted that my dad wasn’t my dad.
About 23 years later, I do a dna test and find out patrol was in fact, correct.
55
At least the dystopia is open source 🥰
We spent the last 11 months developing an AI waifu and she is now OPEN SOURCED!
mekahime.com/openamika
She can:
- respond in 900ms
- see you
- connect to Hermes, nanoMuse, OpenClaw to do real works
- remind you to to exercise
- connect to your smart home devices
Check it out!
1
40
How close to empty should we go before refueling? 🤔
The Sun has only 22 galactic orbits left.
Earth races around the Sun at ~67,000 mph (107,000 km/h), giving us our familiar 365.25-day year and changing seasons.
But the Sun is in motion too, hurtling through the Milky Way at ~514,000 mph (828,000 km/h) on a grand orbit around the galactic center. One complete lap, known as a cosmic year, takes roughly 225–230 million years.
When the Sun finished its most recent galactic orbit, the earliest dinosaurs were just beginning to roam Earth.
Since its birth ~4.6 billion years ago, our star has completed about 20 such orbits.
Stellar models predict the Sun will keep fusing hydrogen in its core for another ~5 billion years before it swells into a red giant and eventually fades into a white dwarf. At its current orbital speed, that leaves roughly 22 more laps around the Milky Way.
Each cosmic year sweeps the entire Solar System tens of thousands of light-years across the galaxy, through dense spiral arms rich with star-forming regions, past ancient globular clusters, and amid countless other stars.
Continents drift, mountains rise and erode, entire species evolve and vanish, all within a tiny fraction of one galactic circuit.
Human civilization, from the first cities to today’s digital age, has existed for less than 0.001% of a single cosmic year.
We are passengers on a star halfway through its ~10-billion-year galactic journey across a 100,000-light-year-wide disk, witnessing just the briefest sliver of one ongoing lap in an unimaginably vast cosmic dance.
41
JEPA has the possibility of being the next gen of AI for sure but not getting hopes up yet
Yann LeCun and Randall Balestriero just Proved the Platonic Representation Hypothesis, and found a better way to achieve it.
The Platonic Representation Hypothesis suggests that AI models trained on different modalities, like images and text, eventually learn similar representations of the world. But this paper shows that how models are trained matters more than simply scaling them up.
So they mathematically show that, under certain conditions, two JEPAs trained completely separately can learn essentially the same underlying representation of the world. One learns from different images of the same scene, while the other learns from different descriptions of that scene, without ever being trained on image-text pairs.
Their approach achieves the same image-text alignment with 42x fewer training samples than next-token prediction. A 109M-parameter text JEPA even outperforms LLMs up to 72B parameters on cross-modal alignment benchmarks.
alphaxiv.org/abs/2610.39302
78
