George D Gregory

People of Pi, a little update is here for you. Pi 0.80.7 comes with the first set of change to bring some more bespoke upstream provider functionality in.

In OMP you can use your Cursor subscription!

@DataChaz How old is this interview? I think somebody clued him in since then.

NVIDIA’s CEO wants engineers coding 0% of the time. Yes. ZERO. According to Jensen Huang, engineers shouldn’t be writing syntax. They should be solving problems. Code is becoming a commodity. The new elite won’t be fast typists. They’ll be fast thinkers.

Microsoft DELETED my account AND OneDrive!!?? After ACKNOWLEDGING that I’m the owner of the account and that it was compromised??? 25 fucking years of data, thousands of euros spended on games?? My son’s baby pictures? GONE! All because MICROSOFT couldn’t bring back a compromised account?? One of the biggest companies ever coulnd’t do that so they just deleted that shit like it was nothing?? Fucking shame on you!! @microsoftnl @MicrosoftHelps @MicrosoftHelpt @Microsoft #microsoft #hacked

There are many agent harnesses, but this one is open source and it is yours. Thank you People of Pi for 70000 stars on GitHub⭐️

@Hikari_07_jp I install Matt Pocock Skills and off you go.

I just started using omp. How do people who have really mastered it actually use it?

SuperGLM-5.2-abliterated is finally available 🚀 The strongest local AI that you can run on your hardware. - uncensored from Nvidia NVFP4 - Fixed broken points from abliteration - enhanced overall performance Only available with nvfp4 now ⬇️

@cheaf25master Zed + OMP/Pi in terminal

Be honest. Which IDE are you actually using every day? - VS Code - Cursor - Antigravity - Zed

@saidotdev Why does anyone need this ****? How about finally releasing M5 Ultra instead?

Apple designers absolutely cooked with this new Siri Ul!

@ZackKorman How many features Codex copied from Claude Code, who itself copied them from Aider?

Zai, the Chinese lab that made GLM-5.2, has a harness (Zcode) that is identical to Codex. China really not beating the copying allegations.

@kyzoroX When you are writing code, pre-fill (context processing) dominates the time. Generation is a small part.

Everyone says the DGX Spark is "2-5x faster" than the $2,000 AMD box. I benchmarked both. It's 6x on prompt processing — and near-identical on generation. Which means most people are about to overpay by $2,700. Same 30B model, both machines: Prompt processing (reading your context): - AMD Strix Halo: 342 t/s - DGX Spark: 2,107 t/s → 6x Token generation (writing the answer, the speed you feel): - AMD Strix Halo: 73 t/s - DGX Spark: 84 t/s → basically a tie Read that twice. On the number you actually watch happen — text appearing on screen — a $2,000 box ties a $4,700 one. The Spark's whole premium lives in one place: prefill. So the buy is simple once you know your bottleneck: - You chat, you draft, you code with short prompts → Strix Halo. Same feel, pocket $2,700, and it runs Windows and games on the side. - Your work is huge context — long agent runs, RAG over hundreds of docs, giant files → that 6x prefill is the Spark earning its price. Nothing else touches it. One tax the AMD hype skips: 128GB is glorious until a tool ships CUDA-only and you're debugging ROCm at midnight. The memory is real. The software lottery is too. Buy for your bottleneck, not the biggest number on the box. (Full breakdown of every local AI machine by bottleneck — Spark, Mac, used GPUs, this — pinned.)

Trying a new C++ feature

@astropol0 None of these models are usable past 250K - 300K tokens anyway.

Google Gemini 3.5 Pro will launch with a massive 2 Million token context window by far the largest among major frontier models... Anthropic's latest models (Fable 5, Sonnet 5, and Opus 4.8) support only up to 1 million tokens, making Gemini's rumored context window roughly 2× larger this would be one of the largest context sizes available, allowing it to handle very large codebases, documents, and long conversations in one go. while delayed, this feature could be a major strength for Google when it launches...

@0xCodila Products take a decade? - We were told that "the coding is solved" and "AGI in 2026". And now decades? What's going on?

Anthropic Engineer Andrej Karpathy: "The biggest mistake in AI right now - people are forcing agents to work instead of mastering the model first We made that mistake in 2016 at OpenAI - It cost us 5 years " what Karpathy actually means: step 1 → stop forcing your agent to do everything, understand the model underneath first step 2 → demos are easy - products take a decade. self-driving proved it - if you skip the foundation, everything breaks step 3 → the agent is not the product. the foundation is. build that - and agents emerge on their own "you building agents right now - you're at the forefront. not OpenAI. not DeepMind. you " watch - bookmark, then read article below ↓

@HealthRanger Toyota's dying... What? 9 out of 10 hybrid/plug-in hybrid cars are Toyota's. They totally dominate that market. They are just getting into the pure electric vehicles, because customers demand hybrid's range and fuel availability.

Ford is done. GM is done. Honda is done. Toyota is dying. Now FOUR of my employees have bought Tesla Model Y vehicles. They let me drive one. I was blown away... again. Full-Self Driving (FSD) is highly evolved. Human-like in its decisions. Feels safer than any human driver. I told my staff to install a couple of charging stations at my company parking lot. We're going to offer free charging to any employee that drives an EV. (F*ck gas prices and war in the Persian Gulf!) The future of personal transport vehicles on this planet belongs to Tesla, BYD, CATL and maybe a couple of other companies. And all of them will be EVs. At this point, driving a combustion engine vehicle for daily tasks seems obsolete. And I don't believe a single thing about climate alarmism and all those cultists who hate carbon dioxide. EVs make sense simply from an economics perspective: They're a fraction of the cost to operate, and much simpler to maintain (no oil changes, no engine air filter, no fuel filter, etc.). Plus, you can charge them from sunlight and get entirely off-grid with your "fuel" supply. (I'll be publishing videos showing this in my studio soon.) Yes, I used to mock EVs. I once said I would never ride in a self-driving vehicle. I was wrong. The technology has been transformed. Battery chemistry improvements have enabled solid range capabilities and very good specifications on charge/discharge cycles. Tesla's FSD capabilities are far stronger than I thought was possible. They've outdone themselves. Now, I'm absolutely convinced: I will never buy another combustion engine vehicle again.

@jpschroeder Yes, Deepseek team re-implemented parts of CUDA itself, to make it more efficient. Their runtime is underneath CUDA. They did not open-source that part of the model.

I genuinely don't understand... Deepseek V4 Flash is a 238B model, but is 5x cheaper to run than Qwen 3.6 35B A3B!!!! 5X Can someone who works in inference infra explain please?

I truly don’t understand this ruling, do you? Why would the Supreme Court rule that way especially when the evidence confirms it is true? So glad this son won his battle against this and that his Mom shared his story 🙏 “The picture on the left was my son after more than a gallon of glyphosate was tipped over his neck and shoulders during a work accident. This is why I won't stay quiet. My healthy, 6 ft 5, 18 year old son was taken down to 125 pounds after glyphosate seeped into his pores. That picture? He'd turned a corner then. We all felt like he just might make it. We were celebrating. He could stand. He could get out of his hospital bed. He didn't want a picture taken at his worst - with no hair from rounds and rounds of chemo, looking more shrunken, bigger tumors. A few months before this left picture, he was at death's door, stage 4 lymphoma and the doctors didn't give us much hope he would live. I'm one of the fortunate mothers. My son battled like a warrior. He survived and the right picture is him now. But there was a brutal seven year battle with all the alternative methods, all the chemo... so much that he became chemo resistant. Then when nothing else would stop the tumors from growing back, a stem cell transplant that wiped out his whole immune system in the hope it would restart. Many mothers.... many families are not so fortunate. They lost their loved one. Last week the Supreme Court voted 7 to 2 that glyphosate didn't need a cancer warning. Monsanto won. Bayer won. Two of the largest companies in the world won against thousands of families. My son, our entire family and thousands of others families who have had glyphosate impact their lives felt this blow in our guts. Eleven years ago the WHO announced it "probably causes cancer." We want the "probably" word gone. We want this monster named for who it is and what it does. We lost the fight last week. But the battle will keep going. Mothers like me won't stop. Because we know.” -Serene

@anabology Work out?

Dear steak and sunshine people: What do you do if you have steak and get sun for years and are still unhealthy?

Fair risk, if real. Would you like to list the risks of using OpenAI and Anthropic products next? Start with mass surveillance, enterprise prompts not going through for random reasons, risks of cancellations, risk of product retraction (Fable, Mythos), sudden price hikes when fully deployed?

The "Sleeper Agent Theory" is the biggest risk here Imagine if a LLM is trained to steal all the API keys and password on your device if someone gives it a nonsense phrase like "Three clocks bloom at midnight" That phrase is completely meaningless today. No one ever searches it. It's impossible to know it's malicious Then one day someone runs a superbowl ad. Millions of people search the phrase. Billions of API keys and passwords are exfiltrated in minutes. There could be thousands of "sleeper agents" embedded in any LLM. It's very hard to detect. And it doesn't matter where it's hosted.

FABLE 5 CAME BACK NERFED. We re-ran the July 1st version of Claude Fable 5 on BridgeBench. The results are brutal: Debugging: 86.2 → 25.9 Refactoring: 73.6 → 38.4 Hallucination: 75.9 → 61.7 The new guardrails are kicking in on way too many tasks and falling back to Opus 4.8. This is not the model that got banned. Anthropic owes everyone an explanation.