#Apache License

[furiouslovepolice]

Mention Source URL, an Attribution PlugIn for WordPress

[jcmarchi]

Alibaba Qwen QwQ-32B: Scaled reinforcement learning showcase

New Post has been published on https://thedigitalinsider.com/alibaba-qwen-qwq-32b-scaled-reinforcement-learning-showcase/

Alibaba Qwen QwQ-32B: Scaled reinforcement learning showcase

The Qwen team at Alibaba has unveiled QwQ-32B, a 32 billion parameter AI model that demonstrates performance rivalling the much larger DeepSeek-R1. This breakthrough highlights the potential of scaling Reinforcement Learning (RL) on robust foundation models.

The Qwen team have successfully integrated agent capabilities into the reasoning model, enabling it to think critically, utilise tools, and adapt its reasoning based on environmental feedback.

“Scaling RL has the potential to enhance model performance beyond conventional pretraining and post-training methods,” the team stated. “Recent studies have demonstrated that RL can significantly improve the reasoning capabilities of models.”

QwQ-32B achieves performance comparable to DeepSeek-R1, which boasts 671 billion parameters (with 37 billion activated), a testament to the effectiveness of RL when applied to robust foundation models pretrained on extensive world knowledge. This remarkable outcome underscores the potential of RL to bridge the gap between model size and performance.

The model has been evaluated across a range of benchmarks, including AIME24, LiveCodeBench, LiveBench, IFEval, and BFCL, designed to assess its mathematical reasoning, coding proficiency, and general problem-solving capabilities.

The results highlight QwQ-32B’s performance in comparison to other leading models, including DeepSeek-R1-Distilled-Qwen-32B, DeepSeek-R1-Distilled-Llama-70B, o1-mini, and the original DeepSeek-R1.

Benchmark results:

AIME24: QwQ-32B achieved 79.5, slightly behind DeepSeek-R1-6718’s 79.8, but significantly ahead of OpenAl-o1-mini’s 63.6 and the distilled models.

LiveCodeBench: QwQ-32B scored 63.4, again closely matched by DeepSeek-R1-6718’s 65.9, and surpassing the distilled models and OpenAl-o1-mini’s 53.8.

LiveBench: QwQ-32B achieved 73.1, with DeepSeek-R1-6718 scoring 71.6, and outperforming the distilled models and OpenAl-o1-mini’s 57.5.

IFEval: QwQ-32B scored 83.9, very close to DeepSeek-R1-6718’s 83.3, and leading the distilled models and OpenAl-o1-mini’s 59.1.

BFCL: QwQ-32B achieved 66.4, with DeepSeek-R1-6718 scoring 62.8, demonstrating a lead over the distilled models and OpenAl-o1-mini’s 49.3.

The Qwen team’s approach involved a cold-start checkpoint and a multi-stage RL process driven by outcome-based rewards. The initial stage focused on scaling RL for math and coding tasks, utilising accuracy verifiers and code execution servers. The second stage expanded to general capabilities, incorporating rewards from general reward models and rule-based verifiers.

“We find that this stage of RL training with a small amount of steps can increase the performance of other general capabilities, such as instruction following, alignment with human preference, and agent performance, without significant performance drop in math and coding,” the team explained.

QwQ-32B is open-weight and available on Hugging Face and ModelScope under the Apache 2.0 license, and is also accessible via Qwen Chat. The Qwen team views this as an initial step in scaling RL to enhance reasoning capabilities and aims to further explore the integration of agents with RL for long-horizon reasoning.

“As we work towards developing the next generation of Qwen, we are confident that combining stronger foundation models with RL powered by scaled computational resources will propel us closer to achieving Artificial General Intelligence (AGI),” the team stated.

See also: Deepgram Nova-3 Medical: AI speech model cuts healthcare transcription errors

Want to learn more about AI and big data from industry leaders? Check out AI & Big Data Expo taking place in Amsterdam, California, and London. The comprehensive event is co-located with other leading events including Intelligent Automation Conference, BlockX, Digital Transformation Week, and Cyber Security & Cloud Expo.

Explore other upcoming enterprise technology events and webinars powered by TechForge here.

[thellawtoknow]

Copyright Protection of Open Source Code

Topic: Understanding Copyright Protection of Open Source Code Open source code is a valuable resource for developers, providing access to a wide range of software and tools that can be used and modified freely. However, many developers may wonder whether open source code is protected by copyright law, and what implications this has for their own use and modification of open source code.…

[whencyclopedia]

Geronimo

Geronimo (Goyahkla, l. c. 1829-1909) was a medicine man and war chief of the Bedonkohe tribe of the Chiricahua Apache nation, best known for his resistance against the encroachment of Mexican and Euro-American settlers and armed forces into Apache territory and as one of the last Native American leaders to surrender to the United States government.

During the Apache Wars (1849-1886), he allied with other leaders such as Cochise (l. c. 1805-1874) and Victorio (l. c. 1825-1880) in attacks on US forces after Apache lands became part of US territories following the Mexican-American War (1846-1848). Between c. 1850 and 1886, Geronimo led raids against villages, outposts, and cattle trains in northern Mexico and southwest US territories, often striking with relatively small bands of warriors against superior numbers and slipping away into the mountains and then back to his homelands in the region of modern-day Arizona and New Mexico.

He surrendered to US authorities three times, but when the terms of his surrender were not honored, he escaped the reservation and returned to launching raids on settlements. He was finally talked into surrendering for good by First Lieutenant Charles B. Gatewood (l. 1853-1896), under the command of General Nelson A. Miles (l. 1839-1925), in 1886. None of the terms stipulated by Miles were honored, but by that time, Geronimo felt he was too old and too tired to continue running. Geronimo's surrender to Gatewood is told accurately, though with some poetic license, in the Hollywood movie Geronimo: An American Legend (1993).

Geronimo was imprisoned at Fort Pickens, Pensacola, Florida, before being moved to Fort Sill, Oklahoma. Toward the end of his life, he became a sensation at the St. Louis World's Fair (1904) and President Theodore Roosevelt's Inaugural Parade (1905) as well as other events. Although one of the stipulations of his surrender was his return to his homelands in Arizona, he was held as a prisoner elsewhere for 23 years before dying in 1909 of pneumonia at Fort Sill.

Name & Youth

His Apache name was Goyahkla ("One Who Yawns"), and, according to some scholars, he acquired the name Geronimo during his campaigns against Mexican troops, who would appeal to Saint Jerome (San Jeronimo in Spanish) for assistance. This was possibly Saint Jerome Emiliani (l. 1486-1537), patron of orphans and abandoned children, not the better-known Saint Jerome of Stridon (l. c. 342-420), translator of the Bible into the Vulgate and patron of translators, scholars, and librarians.

Geronimo was born near Turkey Creek near the Gila River in the region now known as Arizona and New Mexico c. 1825. He was the fourth of eight children and had three brothers and four sisters. In his autobiography, Geronimo: The True Story of America's Most Ferocious Warrior (1906), dictated to S. M. Barrett, Geronimo described his youth:

When a child, my mother taught me the legends of our people; taught me of the sun and sky, the moon and stars, the clouds, and storms. She also taught me to kneel and pray to Usen for strength, health, wisdom, and protection. We never prayed against any person, but if we had aught against any individual, we ourselves took vengeance. We were taught that Usen does not care for the petty quarrels of men. My father had often told me of the brave deeds of our warriors, of the pleasures of the chase, and the glories of the warpath. With my brothers and sisters, I played about my father's home. Sometimes we played at hide-and-seek among the rocks and pines; sometimes we loitered in the shade of the cottonwood trees…When we were old enough to be of real service, we went to the field with our parents; not to play, but to toil.

(12)

After his father died of illness, his mother did not remarry, and Geronimo took her under his care. In 1846, when he was around 17 years old, he was admitted to the Council of Warriors, which meant he could now join in war parties and also marry. He married Alope of the Nedni-Chiricahua tribe, and they would later have three children. Geronimo set up a home for his family near his mother's teepee, and as he says, "we followed the traditions of our fathers and were happy. Three children came to us – children that played, loitered, and worked as I had done" (Barrett, 25). This happy time in Geronimo's life would not last long, however.

Continue reading...

[50calmadeuce]

Ch. 34: The Cabin (R)

Disclaimer: This is a work of fan fiction using characters from the Top Gun: Maverick world, trademarked by Paramount Pictures Corporation. I do not claim ownership of the characters and the world that I am borrowing.

The story and situation I am creating are a work of my imagination and I do not ascribe them to official story canon. This work is for entertainment only and is not a part of the storyline.

I am not profiting financially from the creation and publication of this story, but I do hope it gives you happy thoughts.

These stories are my own, so please do not take them and use them for yourself without my permission. If you see them somewhere else, please let me know. :)

Warning: This chapter has to do with a person being captured by another person. If this is a trigger for you, please don’t read.

A week later, you sat in your office working from home. You had to take some online credits for your Veterinary license. You turned to your left and looked at your wedding pictures. You and Christian’s wedding picture now stood behind yours and Jake’s.

You gazed at the two pictures for a moment, the contrast between them striking. Christian’s wedding photo was filled with youthful joy, a different chapter of your life—one full of love and promise, but ultimately, marked by loss. Now, Jake’s photo was next to it, a new beginning, one that filled your heart with warmth and excitement, but also uncertainty.

It was surreal, how quickly everything had shifted. You never imagined that your path would lead here, but with Jake, you found yourself looking forward to the future instead of lingering on the past.

You lucked out when your photographer captured a great shot of Jake in his dress whites. Honestly, who wouldn’t want a picture of Jake dressed up? That photo was displayed on the right side of your computer, next to the one of Christian in his Apache helicopter.

You couldn’t help but smile every time you glanced at the picture of Jake in his dress whites. There was something about him in uniform, standing so tall and proud, that made your heart flutter. It was more than just the uniform—it was the man beneath it, the one who’d managed to wrap himself so completely into your life in such a short time. His presence in your life had brought with it a sense of security and warmth that you never knew you needed.

On the other side, Christian’s picture in the cockpit of his Apache helicopter evoked a different kind of feeling—one of reverence for his dedication and the memory of all that he had been. The contrast between the two photos spoke volumes: two men who had shaped your life in their own unique ways, but both with a deep commitment to serving something greater than themselves. You felt an overwhelming sense of peace, knowing that, even though life had taken unexpected turns, you had found love again in the most unexpected of places.

You continued sipping your coffee and reading an online chapter when your phone dinged. You glanced at it:

Possible elk kill. Need you to confirm. Sending location.

You sighed. Guess your certification would have to wait. You stood up and headed to your bedroom to get dressed.

An hour later, you drove down a gravel road that seemed to have been untouched for quite some time. According to the GPS, you were headed in the right direction.

Suddenly, a slightly overgrown clearing appeared, revealing a dilapidated log cabin, and a familiar truck came into view.

It was Scott’s.

You parked your truck next to his and turned off the engine, glancing around at the terrain and the cabin. The only thing that stood out was the smoke rising from the chimney.

Unbuckling your seatbelt, you carefully leaned forward to check if your boot knife was secure in your right boot. Satisfied, you took the keys out of the ignition, cautiously opened the truck door, and stepped out, closing it quietly behind you.

“Scott?” you called out.

To your right, startled birds flew out of a tree, making you turn to the noise, placing your hand over your Glock, getting ready to use it if needed.

Then the door of the cabin opens, and Scott walks out. You relax only slightly.

“Y/N. The elk, or what’s left of it, is this way.” He walks down the steps and then starts walking to his left, your right. The same way the birds flying out of the tree came from.

You watch him and he stops. “This way.”

You look at him questionably. Something was telling you something wasn’t right, so you slowly start to follow him, and he continued walking.

“Whose property is this?” You ask as you catch up to him but keep some distance.

“Some family in New York. They saw the elk on their camera.”

You stop in your tracks, looking at Scott, a growing sense of unease creeping up your spine. "I didn’t see any cameras when I pulled in."

Scott pauses, glancing over his shoulder at you. For a brief moment, his expression tightens, and then he gives you a casual smile. "Probably just missed them. They're pretty well hidden."

You nod, but something still doesn’t sit right. The hairs on the back of your neck stand up as you take another look around the clearing, eyes scanning the perimeter. There’s no sign of cameras, no obvious surveillance equipment, just the woods stretching out around the cabin. The smell of fresh elk blood is in the air, but it doesn’t feel like you’re out here for a simple confirmation of a kill.

You slowly take a step forward, keeping an eye on Scott. “And why did you call me out here? For a confirmation?”

Scott glances at you over his shoulder again, his pace slowing. “Just making sure we’re on the same page. I thought you’d want to take a look yourself. That’s all.”

But something about his tone feels off. Something's wrong here, and you can feel it in your gut.

Your heart rate picks up, and you fight to keep your voice steady. "Just making sure, huh?" You take a careful step forward, but your instincts are screaming for you to step back, to turn and leave before whatever this is goes any further.

Scott doesn’t turn around, but you can tell he's aware of the shift in your demeanor. His shoulders tense, his jaw tightens, and that easy smile of his falters for a brief moment before he turns and faces you fully, the mask of casual confidence slipping just enough to reveal something more calculating beneath.

“Listen,” he says, voice low but sharp, “we’re all just trying to keep things under control. I wouldn’t have called you out here if I didn’t think it was important.”

You lock eyes with him, every muscle in your body now on high alert. "Important for who, Scott?" you ask, a quiet challenge in your voice.

He opens his mouth to speak but is interrupted by a sudden rustling in the brush nearby. Your body stiffens, every muscle instinctively coiling like a spring, ready to react. You glance in the direction of the noise, but before you can process it fully, Scott moves between you and the sound, his posture shifting into something more protective... or perhaps more calculated.

"Stay close," he mutters, his voice tense. You don't need to be told twice. You step back, your hand instinctively moving to the gun at your hip. Something isn't right, and now you can feel the weight of that unease pressing down on you.

Scott seems to be waiting for something, his eyes darting around, scanning the perimeter as if anticipating something. You swallow, trying to steady your breathing. What the hell is going on?

"I’m sorry, Y/N," he said.

Before you had a chance to react, Scott sprayed something into your face, and everything went dark.

Tags: @smoothdogsgirl @alwayshave-faith @devil-angel-winchester @khouse712 @illisea @hookslove1592 @tgmreader @juliemaruaderfan @djs8891

[debian-official]

speaking of,

if you say "wheres CC-whatever" go make your own poll

[kalpeavaris]

Do you have any recommendations for alternative writing programs instead of Google Docs? I wasn't aware of the scraping issue until just now ;;-;;

It honestly depends in what you're expecting from your writing program, do you require a full-blown program you can also sort notes, images, worlds, character sketches and more? Or is a more simplistic text editor sufficient? Every author's different so it honestly comes down to personal preference :D

For simpler programs I can honestly recommend LibreOffice & Apache OpenOffice, they're free as far as I'm concerned (has been a while since I used them) and fullfill the basic needs for a writer.

If you're looking into a more in-depth program I can recommend Scrivener, though that one costs money - around 50-60 bucks, my price was in EUR so not sure how much it'd be in USD. It's really detailed and since my fanfictions/stories usually exceed 100-200 pages each I found myself much more immersed in a very in-depth program meant for bookwriting rather than a basic text editor.

Otherwise Wikipedia has a pretty good comparison of available text editors/programs on hand where you can check which ones offer which features and if they're free or require you to purchase a license: https://en.wikipedia.org/wiki/Comparison_of_text_editors

[eblu3]

experiment. you can now send notifications to my phone. just keep it clean please

[sztupy]

In an effort to change that dynamic, we are announcing today the general availability of a foundational open data set, Foursquare Open Source Places (“FSQ OS Places”). This base layer of 100mm+ global places of interest (“POI”) includes 22 core attributes (see schema here) that will be updated monthly and available for commercial use under the Apache 2.0 license framework.

Nocsak

[celticcrossanon]

Celta, if you look at photos on the balcony of Buckingham Palace for the 100th anniversary of the RAF in 2018— you'll see William is wearing a gold wing insignia, which means he is a fully qualified pilot. Harry's only wearing a silver wing insignia, which are his Provisional Pilot’s Wings (a.k.a. a driver's learner permit). Proof that Harry isn't a qualified pilot nor did he bother to meet the required hours at any time to become a fully licensed pilot before leaving the Army in 2015.

Hi Nonny,

Yes, the insignia differs even though Prince William and Harry are wearing the same uniform, as per below. I am glad the media have finally stopped the nonsense about Harry being an Apache pilot and are referring to him correctly as a gunner or gunner co-pilot. As the retired colonel said, Harry was never number one in the aircraft, he was always number two (and there is no shame for that unless you pretend you were number one, which Harry seems to do not all the time, but on occasion).

[mememanufactorum]

Badger’s Best of 2023 sentence starters

* FEEL FREE TO SHARE AS YOU PLEASE, NO CREDIT NEEDED. CHANGE PRONOUNS OR ANYTHING ELSE AS DESIRED

All lines are from this video created by TheRussianBadger.

"I ACTUALLY EARNED ONE, MOTHERFUCKERS!"

"Those noises that were coming out of you were inhuman."

"You ever had a hotdog burger before?"

"You did NOT just come up with that word."

"I need to know if this was a riff or if this was an actual meal."

"I heard the word 'hotdurger' unprovoked."

"Dudes with nut allergies when I hit them in the head with a brick."

"YOU DIDN'T JUJU ON THE FUCKIN' BEAT."

"I don't misinform. I just lie."

"Did you just punch someone for all their coins?"

"I don't know, just blow 'em all up, I don't care."

"I just fucken hate you."

"STOP BLINDING ME, YOU ASSHOLE! I CAN'T SEE, YOU GOBLIN!"

"To the charge of wire fraud, you are pleading 'nuh-uh'?"

"Your honor, shut the fuck up. You wasn't even there."

"This conversation sounds like four raccoons with internet access."

"You wanna know how I got these GAINS?"

"I was driving through upstate New York and I saw a Tesla with the license plate 'I'M HIM'."

"That license plate made me laugh so hard that I walked up to his window and put a 12-gauge slug in his chest."

"You got me fucked up bro, I can't believe you would question if I'm real."

"Here's a picture of my nuts."

"Those are gonna be my dying words to my wife: I just want you to know… PS3 has no games."

"Chimichangas are a CIA psyop."

"If you put me in the cockpit of an apache I will Kevin Gates, put my hand on the dashboard, and start it."

"Boy I love having something with none of the same consistency as anything else in my sandwich in my sandwich."

"Dude I definitely love biting into my sandwich and then leaving with an entire pickle slice in my mouth."

"Own a musket for home defense since that's what the founding fathers intended."

"I have to resort to the cannon mounted at the top of the stairs loaded with grapeshot. Tally ho, lads!"

"Well it's just straight up racism, and it's not even like an occasional racism, it's like, this is full blast firehose racism."

"It's the floodgate of racism! The Big Gulp of racism!"

"This shit will turn your pacemaker off."

"I point blanked that shit with a panzerfaust."

"Me going to Arby's after losing a $50,000 Marvel vs Capcom tournament."

"Me walking to the fridge to get my five day old caesar salad."

"Fresh caesar salad, already not a good start. Five days, dog."

"How does that predator missile work? Oh, you just go NYOOOOOOM."

"This Nyquil beatin' my ass, that is not THAT funny but, like, I can't stop laughing!"

"Y'all just verbally buzzered that man."

"I stole your girl, I stole your whip, I stole your shoes."

"You cannot land a KC-135 in a Kroger parking lot."

"As someone who lives in Tennessee, you can land a KC-135 in a Kroger parking lot."

"That's how I'm going to describe the size of our parking lots to Europeans without internet connections. We can land that in our parking lots."

"I call that my main menu tax."

"Bro, I can't hail a cab in Detroit for shit, bro."

"First bullet, Toyota Tacoma be like 'I ain't hear NOTHING. Y'all hear something?' Second bullet? Legalize nuclear bombs."

"Your voice literally has to wait in line to be heard."

"I'm gonna bomb your trailer park."

"Don't take advice from the dead guys."

"Smoking on that diabolical arch-necromancer pack. Those who don't ball would do well to steer clear."

"Do you know the word 'whermst'?"

"It's like where and for what purpose and why. Location, reason, background context in one word: Whermst."

"Did he just prefire me? Bro, go to jail."

"That's your first option for recourse?"

"Alcatraz, we ain't talking county jail. You're getting in there with the dementors."

"Stop calling the 3D avatar mommy."

"How do they fit this many flares in an airplane? It makes no sense. It's like a clown car but for fireworks."

"I'M SCREAMING ABOUT IT MOTHERFUCKER, STOP!"

"Hey what's up guys? I just bought a 1911 at a Red Lobster parking lot, AMA."

"Just kill me. Just take me to heaven. Just… Take me out of this reality."

"Heaven? BITCH, YOU GOING TO HELL!"

"Hey, fuckin' imagine getting friendly fired by a .50 BMG. Imagine."

"My client pleads oopsie-daisy."

"I'm sorry that your dog is not going to college now."

"Ay you ain't on your grind, son. You ain't on your bag."

"No one's Batman impression is bad."

"You sound like you're in an alley with a trench coat, what the fuck?"

"Oh my God, his Scooby-Doo villain is coming out again."

"Are you repairing our conversation?"

"Why is 'slime' such a funny yet affectionate nickname?"

"Get the fuck out of our shower."

"Why can't we just share the shower?"

"Enemy. Man. 300 meters. North. Fast. Fast. Fast."

"Fun fact: The TSA allows you to bring a live lobster through security."

"I myself have brought 432 lobsters through security."

"THAT'S THE FOURTH TIME YOU'VE SHOT ME!"

"SHUT UP! YOU JUST HAPPEN TO BE WHERE MY BULLETS ARE!"

"All units, be advised: My stummy hurt."

"Homie got the dog in him with that one."

"Pulled pork? Yeah I cranked my hog today too."

"How blessed are we that I can just log on to YouTube and the first video I see is 'Master Chief teaches you how to change the oil on your 2006 Nissan Murano'?"

"That went from 'funny' to 'demonitized'."

"If your state has 90 degree corners, you probably eat corn syrup on your pancakes."

"Why do you always say 'theoretically' and it's not at all theoretical?"

"You have the world's WORST EVERYTHING."

"My boy got the object permanence of a frog."

"That boy cooked the most rare steak."

"I gotta use the bathroom or something, bro. I gotta go to college or something. I can't be with these motherfuckers."

"He went behind the tree and my brain was like 'WHERE'D HE GO?'"

"Somebody buy me a stat reset, PLEASE!"

"You should not be legally allowed to commit crimes if you're listening to Lynyrd Skynyrd."

"I'm on my Super Mario Sunshine shit."

"Are you barking at me?"

"You might wanna be a LITTLE shidded right now."

"I'd trust Gengar with my kid."

"I didn't know he was chill like that."

"No. We are not putting a controller around somebody's neck and twisting it. It's a wireless controller, you can't even do that."

"And 45 is just a caliber."

"Ranch was made by California to keep the Midwest fat because they're scared of our power."

"I refuse to believe that Kranch is real."

"Alignment charts are for the governable. I grow corn in my yard."

"Tell me the name of God you fungal piece of shit."

"I'm pretty sure that was the most sacrilegious shit I've heard in my life."

"I will pass that to the higher ups – parentheses: I do not give a shit."

"This is getting a little too fast for my brain."

"You fuckers are at a pie eating contest and I'm just like, nah son. Free pie."

"I'm about to hit 'em with the Glock-no-jutsu, on God, bro."

"Regretting a free purchase is crazy."

"THEY'RE JUST POLYGONS!"

"I've had people call me things that I wouldn't even dare say to myself."

"Take five 5-Hour Energies and enter the forbidden hour of the day."

"Those responses do not surprise me at all. I definitely expected that kind of language."

"Bro, it's goof-a-clock right now."

"The moon already isn't real."

"You think I can't kill a fuckin' banana?"

"That was a little too much rage for a potassium transportation device. I didn't mean it. You full of electrolytes."

"I'm gonna eat pizza because I like the sauce on the pizza with the cheese on the pizza."

"I could not have killed him any harder."

"Don't make me make you say some out of pocket shit."

"I've been saying out of pocket shit all day."

"By sheer artillery alone, we should have tunneled our way to Atlantis by now."

"Yo, I don't know the Tom & Jerry lore, fuck you!"

"What if you wanted to go to heaven but God said to you, 'WE'RE GONNA TRY THIS WEEK'S CRUMBL COOKIE MENU'?"

"I CAN'T HEAR YOU OVER THE TINNITUS, WHAT?"

"Is this like punching someone in the dark? Is it like a legal loophole?"

"There's only one of me in all the world. I am one in a krillion."

"If you're a chest sleeper, you're just a fuckin' psychopath, alright?"

[jcmarchi]

Karthik Ranganathan, Co-Founder and Co-CEO of Yugabyte – Interview Series

New Post has been published on https://thedigitalinsider.com/karthik-ranganathan-co-founder-and-co-ceo-of-yugabyte-interview-series/

Karthik Ranganathan, Co-Founder and Co-CEO of Yugabyte – Interview Series

Karthik Ranganathan is co-founder and co-CEO of Yugabyte, the company behind YugabyteDB, the open-source, high-performance distributed PostgreSQL database. Karthik is a seasoned data expert and former Facebook engineer who founded Yugabyte alongside two of his Facebook colleagues to revolutionize distributed databases.

What inspired you to co-found Yugabyte, and what gaps in the market did you see that led you to create YugabyteDB?

My co-founders, Kannan Muthukkaruppan, Mikhail Bautin, and I, founded Yugabyte in 2016. As former engineers at Meta (then called Facebook), we helped build popular databases including Apache Cassandra, HBase, and RocksDB – as well as running some of these databases as managed services for internal workloads.

We created YugabyteDB because we saw a gap in the market for cloud-native transactional databases for business-critical applications. We built YugabyteDB to cater to the needs of organizations transitioning from on-premises to cloud-native operations and combined the strengths of non-relational databases with the scalability and resilience of cloud-native architectures. While building Cassandra and HBase at Facebook (which was instrumental in addressing Facebook’s significant scaling needs), we saw the rise of microservices, containerization, high availability, geographic distribution, and Application Programming Interfaces (API). We also recognized the impact that open-source technologies have in advancing the industry.

People often think of the transactional database market as crowded. While this has traditionally been true, today Postgres has become the default API for cloud-native transactional databases. Increasingly, cloud-native databases are choosing to support the Postgres protocol, which has been ingrained into the fabric of YugabyteDB, making it the most Postgres-compatible database on the market. YugabyteDB retains the power and familiarity of PostgreSQL while evolving it to an enterprise-grade distributed database suitable for modern cloud-native applications. YugabyteDB allows enterprises to efficiently build and scale systems using familiar SQL models.

How did your experiences at Facebook influence your vision for the company?

In 2007, I was considering whether to join a small but growing company–Facebook. At the time, the site had about 30 to 40 million users. I thought it might double in size, but I couldn’t have been more wrong! During my over five years at Facebook, the user base grew to 2 billion. What attracted me to the company was its culture of innovation and boldness, encouraging people to “fail fast” to catalyze innovation.

Facebook grew so large that the technical and intellectual challenges I craved were no longer present. For many years I had aspired to start my own company and tackle problems facing the common user–this led me to co-create Yugabyte.

Our mission is to simplify cloud-native applications, focusing on three essential features crucial for modern development:

First, applications must be continuously available, ensuring uptime regardless of backups or failures, especially when running on commodity hardware in the cloud.

Second, the ability to scale on demand is crucial, allowing developers to build and release quickly without the delay of ordering hardware.

Third, with numerous data centers now easily accessible, replicating data across regions becomes vital for reliability and performance.

These three elements empower developers by providing the agility and freedom they need to innovate, without being constrained by infrastructure limitations.

Could you share the journey from Yugabyte’s inception in 2016 to its current status as a leader in distributed SQL databases? What were some key milestones?

At Facebook, I often talked with developers who needed specific features, like secondary indexes on SQL databases or occasional multi-node transactions. Unfortunately, the answer was usually “no,” because existing systems weren’t designed for those requirements.

Today, we are experiencing a shift towards cloud-native transactional applications that need to address scale and availability. Traditional databases simply can’t meet these needs. Modern businesses require relational databases that operate in the cloud and offer the three essential features: high availability, scalability, and geographic distribution, while still supporting SQL capabilities. These are the pillars on which we built YugabyteDB and the database challenges we’re focused on solving.

In February 2016, the founders began developing YugabyteDB, a global-scale distributed SQL database designed for cloud-native transactional applications. In July 2019, we made an unprecedented announcement and released our previously commercial features as open source. This reaffirmed our commitment to open-source principles and officially launched YugabyteDB as a fully open-source relational database management system (RDBMS) under an Apache 2.0 license.

The latest version of YugabyteDB (unveiled in September) features enhanced Postgres compatibility. It includes an Adaptive Cost-Based Optimizer (CBO) that optimizes query plans for large-scale, multi-region applications, and Smart Data Distribution that automatically determines whether to store tables together for lower latency, or to shard and distribute data for greater scalability. These enhancements allow developers to run their PostgreSQL applications on YugabyteDB efficiently and scale without the need for trade-offs or complex migrations.

YugabyteDB is known for its compatibility with PostgreSQL and its Cassandra-inspired API. How does this multi-API approach benefit developers and enterprises?

YugabyteDB’s multi-API approach benefits developers and enterprises by combining the strengths of a high-performance SQL database with the flexibility needed for global, internet-scale applications.

It supports scale-out RDBMS and high-volume Online Transaction Processing (OLTP) workloads, while maintaining low query latency and exceptional resilience. Compatibility with PostgreSQL allows for seamless lift-and-shift modernization of existing Postgres applications, requiring minimal changes.

In the latest version of the distributed database platform, released in September 2024, features like the Adaptive CBO and Smart Data Distribution enhance performance by optimizing query plans and automatically managing data placement. This allows developers to achieve low latency and high scalability without compromise, making YugabyteDB ideal for rapidly growing, cloud-native applications that require reliable data management.

AI is increasingly being integrated into database systems. How is Yugabyte leveraging AI to enhance the performance, scalability, and security of its SQL systems?

We are leveraging AI to enhance our distributed SQL database by addressing performance and migration challenges. Our upcoming Performance Copilot, an enhancement to our Performance Advisor, will simplify troubleshooting by analyzing query patterns, detecting anomalies, and providing real-time recommendations to troubleshoot database performance issues.

We are also integrating AI into YugabyteDB Voyager, our database migration tool that simplifies migrations from PostgreSQL, MySQL, Oracle, and other cloud databases to YugabyteDB. We aim to streamline transitions from legacy systems by automating schema conversion, SQL translation, and data transformation, with proactive compatibility checks. These innovations focus on making YugabyteDB smarter, more efficient, and easier for modern, distributed applications to use.

What are the key advantages of using an open-source SQL system like YugabyteDB in cloud-native applications compared to traditional proprietary databases?

Transparency, flexibility, and robust community support are key advantages when using an open-source SQL system like YugabyteDB in cloud-native applications. When we launched YugabyteDB, we recognized the skepticism surrounding open-source models. We engaged with users, who expressed a strong preference for a fully open database to trust with their critical data.

We initially ran on an open-core model, but rapidly realized it needed to be a completely open solution. Developers increasingly turn to PostgreSQL as a logical Oracle alternative, but PostgreSQL was not built for dynamic cloud platforms. YugabyteDB fills this gap by supporting PostgreSQL’s feature depth for modern cloud infrastructures. By being 100% open source, we remove roadblocks to adoption.

This makes us very attractive to developers building business-critical applications and to operations engineers running them on cloud-native platforms. Our focus is on creating a database that is not only open, but also easy to use and compatible with PostgreSQL, which remains a developer favorite due to its mature feature set and powerful extensions.

The demand for scalable and adaptable SQL solutions is growing. What trends are you observing in the enterprise database market, and how is Yugabyte positioned to meet these demands?

Larger scale in enterprise databases often leads to increased failure rates, especially as organizations deal with expanded footprints and higher data volumes. Key trends shaping the database landscape include the adoption of DBaaS, and a shift back from public cloud to private cloud environments. Additionally, the integration of generative AI brings opportunities and challenges, requiring automation and performance optimization to manage the growing data load.

Organizations are increasingly turning to DBaaS to streamline operations, despite initial concerns about control and security. This approach improves efficiency across various infrastructures, while the focus on private cloud solutions helps businesses reduce costs and enhance scalability for their workloads.

YugabyteDB addresses these evolving demands by combining the strengths of relational databases with the scalability of cloud-native architectures. Features like Smart Data Distribution and an Adaptive CBO, enhance performance and support a large number of database objects. This makes it a competitive choice for running a wide range of applications.

Furthermore, YugabyteDB allows enterprises to migrate their PostgreSQL applications while maintaining similar performance levels, crucial for modern workloads. Our commitment to open-source development encourages community involvement and provides flexibility for customers who want to avoid vendor lock-in.

With the rise of edge computing and IoT, how does YugabyteDB address the challenges posed by these technologies, particularly regarding data distribution and latency?

YugabyteDB’s distributed SQL architecture is designed to meet the challenges posed by the rise of edge computing and IoT by providing a scalable and resilient data layer that can operate seamlessly in both cloud and edge contexts. Its ability to automatically shard and replicate data ensures efficient distribution, enabling quick access and real-time processing. This minimizes latency, allowing applications to respond swiftly to user interactions and data changes.

By offering the flexibility to adapt configurations based on specific application requirements, YugabyteDB ensures that enterprises can effectively manage their data needs as they evolve in an increasingly decentralized landscape.

As Co-CEO, how do you balance the dual roles of leading technological innovation and managing company growth?

Our company aims to simplify cloud-native applications, compelling me to stay on top of technology trends, such as generative AI and context switches. Following innovation demands curiosity, a desire to make an impact, and a commitment to continuous learning.

Balancing technological innovation and company growth is fundamentally about scaling–whether it’s scaling systems or scaling impact. In distributed databases, we focus on building technologies that scale performance, handle massive workloads, and ensure high availability across a global infrastructure. Similarly, scaling Yugabyte means growing our customer base, enhancing community engagement, and expanding our ecosystem–while maintaining operational excellence.

All this requires a disciplined approach to performance and efficiency.

Technically, we optimize query execution, reduce latency, and improve system throughput; organizationally, we streamline processes, scale teams, and enhance cross-functional collaboration. In both cases, success comes from empowering teams with the right tools, insights, and processes to make smart, data-driven decisions.

How do you see the role of distributed SQL databases evolving in the next 5-10 years, particularly in the context of AI and machine learning?

In the next few years, distributed SQL databases will evolve to handle complex data analysis, enabling users to make predictions and detect anomalies with minimal technical expertise. There is an immense amount of database specialization in the context of AI and machine learning, but that is not sustainable. Databases will need to evolve to meet the demands of AI. This is why we’re iterating and enhancing capabilities on top of pgvector, ensuring developers can use Yugabyte for their AI database needs.

Additionally, we can expect an ongoing commitment to open source in AI development. Five years ago, we made YugabyteDB fully open source under the Apache 2.0 license, reinforcing our dedication to an open-source framework and proactively building our open-source community.

Thank you for all of your detailed responses, readers who wish to learn more should visit YugabyteDB.

[omgmaxsmith-blog]

Can Open Source Integration Services Speed Up Response Time in Legacy Systems?

Legacy systems are still a key part of essential business operations in industries like banking, logistics, telecom, and manufacturing. However, as these systems get older, they become less efficient—slowing down processes, creating isolated data, and driving up maintenance costs. To stay competitive, many companies are looking for ways to modernize without fully replacing their existing systems. One effective solution is open-source integration, which is already delivering clear business results.

Why Faster Response Time Matters

System response time has a direct impact on business performance. According to a 2024 IDC report, improving system response by just 1.5 seconds led to a 22% increase in user productivity and a 16% rise in transaction completion rates. This means increased revenue, customer satisfaction as well as scalability in industries where time is of great essence.

Open-source integration is prominent in this case. It can minimize latency, enhance data flow and make process automation easier by allowing easier communication between legacy systems and more modern applications. This makes the systems more responsive and quick.

Key Business Benefits of Open-Source Integration

Lower Operational Costs

Open-source tools like Apache Camel and Mule eliminate the need for costly software licenses. A 2024 study by Red Hat showed that companies using open-source integration reduced their IT operating costs by up to 30% within the first year.

Real-Time Data Processing

Traditional legacy systems often depend on delayed, batch-processing methods. With open-source platforms using event-driven tools such as Kafka and RabbitMQ, businesses can achieve real-time messaging and decision-making—improving responsiveness in areas like order fulfillment and inventory updates.

Faster Deployment Cycles: Open-source integration supports modular, container-based deployment. The 2025 GitHub Developer Report found that organizations using containerized open-source integrations shortened deployment times by 43% on average. This accelerates updates and allows faster rollout of new services.

Scalable Integration Without Major Overhauls

Open-source frameworks allow businesses to scale specific parts of their integration stack without modifying the core legacy systems. This flexibility enables growth and upgrades without downtime or the cost of a full system rebuild.

Industry Use Cases with High Impact

Banking

Integrating open-source solutions enhances transaction processing speed and improves fraud detection by linking legacy banking systems with modern analytics tools.

Telecom

Customer service becomes more responsive by synchronizing data across CRM, billing, and support systems in real time.

Manufacturing

Real-time integration with ERP platforms improves production tracking and inventory visibility across multiple facilities.

Why Organizations Outsource Open-Source Integration

Most internal IT teams lack skills and do not have sufficient resources to manage open-source integration in a secure and efficient manner. Businesses can also guarantee trouble-free setup and support as well as improved system performance by outsourcing to established providers. Top open-source integration service providers like Suma Soft, Red Hat Integration, Talend, TIBCO (Flogo Project), and Hitachi Vantara offer customized solutions. These help improve system speed, simplify daily operations, and support digital upgrades—without the high cost of replacing existing systems.

[jbird-the-manwich]

i trained an AI for writing incantations.

You can get the model, to run on your own hardware, under the cut. it is free. finetuning took about 3 hours with PEFT on a single gpu. It's also uncensored. Check it out:

The model requires a framework that can run ggufs, like gpt4all, Text-generation-webui, or similar. These are free and very easy to install.

You can find the model itself as a gguf file here:

About:

it turned out functional enough at this one (fairly linguistically complex) task and is unique enough that I figured I'd release it in case anyone wants the bot. It would be pretty funny in a discord. It's slightly overfit to the concept of magic, due to having such a small and intensely focused dataset.

Model is based on Gemma 2, is small, really fast, very funny, not good, dumb as a stump, (but multingual) and is abliterated. Not recommended for any purpose. It is however Apache 2.0 Licensed, so you can sell its output in books, modify it, re-release it, distill it into new datasets, whatever.

it's finetuned on a very small, very barebones dataset of 400 instructions to teach it to craft incantations based on user supplied intents. It has no custom knowledge of correspondence or spells in this release, it's one thing is writing incantations (and outputting them in UNIX strfile/fortune source format, if told to, that's it's other one thing).

magic related questions will cause this particular model to give very generic and internetty, "set your intention for Abundance" type responses. It also exhibits a failure mode where it warns the user that stuff its OG training advises against, like making negative statements about public figures, can attract malevolent entities, so that's very fun.

the model may get stuck repeating itself, (as they do) but takes instruction to write new incantations well, and occasionally spins up a clever rhyme. I'd recommend trying it with lots of different temperature settings to alter its creativity. it can also be guided concerning style and tone.

The model retains Gemma 2's multilingual output, choosing randomly to output latin about 40% of the time. Lots of missed rhymes, imperfect rhythm structures, and etc in english, but about one out of every three generated incantations is close enough to something you'd see in a book that I figure'd I'd release it to the wild anyway.

it is, however, NOT intended for kids or for use as any kind of advice machine; abliteration erodes the models refusal mechanism, resulting in a permanent jailbreak, more or less. This is kinda necessary for the use case (most pre-aligned LLMs will not discuss hexes. I tell people this is because computers belieb in magic.), but it does rend the models safeguards pretty much absent. Model is also *quite* small, at around 2.6 billion parameters, and a touch overfit for the purpose, so it's pretty damn stupid, and dangerous, and will happily advise very stupid shit or give very wrong answers if asked questions, so all standard concerns apply and doubly so with this model, and particularly because this one is so small and is abliterated. it will happily "Yes, and" pretty much any manner of question, which is hilarious, but definitely not a voice of reason:

it may make mistakes in parsing instructions altogether, reversing criteria, getting words mixed up, and sometimes failing to rhyme. It is however pretty small, at 2 gigs, and very fast, and runs well on shitty hardware. It should also fit on edge devices like smartphones or a decent SBC.

for larger / smarter models, the incantation generation function is approximated in a few-shot as a TavernAI card here:

If you use this model, please consider posting any particularly "good" (or funny) incantations it generates, so that I can refine the dataset.

[vintagelasvegas]

Apache Hotel / Horseshoe Club / Binion’s Gambling Hall

Apache Hotel, Fremont & 2nd, right around the time of opening in March 1932. No signs on the building. Photo by Larry Ullom.

Apache hotel and various clubs:

‘05: Property bought by W.R. Thomas, begins building a hotel in ‘10, abandons the project, leaving exposed open basement through the 20s.

‘31: P.O. Silvagni buys the property, begins building. A.L. Worswick, architect.

‘32: Apache Hotel opens 3/19/32. Hotel leased to R.R. Russell for 15 years. Silvagni opens a series of clubs in the basement in ‘32-34 - Apache Indian Village, Kiva Club, and ‘Pache Club. Apache Casino, a separate business, opens on the ground floor in Oct. by Pop Houskey, T. Thebo.

‘41: New Western Casino on the ground floor in Jul.; closed Spring ‘42.

‘42: Forester Jewelers on the ground floor (10/1/42-7/28/44).

‘45: S.S. Rex Club on the ground floor, opened 3/3/45. Partners include T. Cornero, Silvagni, W. Alderman. Sign by Nevada Outdoor Advertising. City orders partners including owner/landlord Silvagni out in Jun; G. McAfee licensed to take over the club in Jul. The "S.S." was dropped from the club name on paper post-Cornero, but the signs never changed.

‘46: Rex Club sold to M. Bernstein in Feb; M. Sedway as manager; casino becomes Eldorado Club in Aug, Sedway continues as manager.

‘47: Apache Hotel rooftop sign installed (c. late ‘47). Arthur Rozen, Ed Moss take over management at Eldorado Club.

‘50: Eldorado closed in Dec. The club reopened as Clover Club circa Jan.-Feb ‘51 with Eldorado sign in place.

‘51: May, Benny Binion buys and closes Clover Club. Eldorado sign removed from marquee that month as renovations to the club begin.

Horseshoe club:

‘51: Horseshoe Club opens on the ground floor 8/14/51.

‘52: “Benny Binion” name added to the Horseshoe sign c. May, after Binion is licensed.

‘53: Binion loses license in Dec. after federal conviction; casino is owned, on paper, by new partners Joe W. Brown, W. Dorsett, “Doby Doc” Caudill; Binion name removed from the sign in late ‘53 or early ‘54.

‘54: Dec. 10, Grand opening Horseshoe Club's new restaurant and bingo room in a new building on South 2nd St, and the Horseshoe's Million Dollar Display. Apache Hotel Annex occupies the second floor of the new building. An alley separates the main building and the annex.

'57: Mar. 17, Binion released from prison, returns to Las Vegas.

‘58: Horseshoe Club leases Apache Hotel. Rooftop sign removed, replaced with a sign that says "Hotel," and the Apache name retired until 2019. Brown’s stake sold to Ed Levinson & Fremont Hotel partners, Levinson becomes president of Horseshoe.

‘59: Million dollar display removed, currency belonging to Joe W. Brown estate.

‘60: Jack Binion, age 22, approved as 2 ½ owner of Horseshoe

‘61: Neon facade installed on the building and expansion onto former Boulder Club property.

‘62: Parking garage addition, McAllister & Wagner, architect.

‘64: Binion family becomes sole shareholders of the Horseshoe. Main building and annex joined, removing the alley at unknown date in mid-late 60s.

‘69: Million Dollar Display re-introduced.

‘70: World Series of Poker begins.

‘88: Horseshoe buys The Mint in Jun; expands into the club in Jul; “Mint” lettering immediately removed from the signs; signs entirely replaced throughout Fall ‘88-early ‘89. Benny Binion statue sculpted by Deborah Copenhaver, at Casino Center & Ogden, unveiled in Nov.

‘90: Culinary Local 226 and Bartenders Local 165 union workers at Binion's Horseshoe go on strike, Jan. 27 to Nov. 7.

‘98: Becky Binion Behnen sole owner of Horseshoe.

‘99: Million Dollar Display sold, removed.

‘04: Horseshoe closed after cash seizure, 1/9/2004. Sold to Harrah’s. Harrah’s retains Horseshoe brand and World Series of Poker, sells remaining assets to MTR Gaming.

Binion’s Gambling Hall:

‘05: Hotel-casino rebranded Binion’s Gambling Hall & Hotel in Mar.

‘08: Casino and hotel lease sold to TLC Gaming.

‘09: Hotels closed (former Mint tower and Apache).

‘19: Older hotel reopens as Apache Hotel.

Photos from the collection at Nevada State Museum, Las Vegas. Sources include: D. Swanson. Blood aces: the wild ride of Benny Binion, the gangster who created Vegas poker. Viking, 2014.

Below: circa 1932, view on 2nd Street, main hotel entrance.