SSSloppyDev

@TheAhmadOsman How many tps you get?

Running DeepSeek V4 Flash on this right now That’s Opus 4.6 quality at home without limits, degraded quality, or giving away any of my private data 6+ years old hardware btw

@officer_secret Sad to tell you this but Flock cameras can see through ir light. It doesn't block their vision at all as they have Advanced Auto-Exposure Optimization

In 2021, I discovered and recreated a mind-blowing physical privacy project: a custom baseball cap fitted with infrared diodes that can hide you from surveillance cameras (hello, Flock!). It was an incredible experience, but unfortunately there weren’t enough buyers. I just stumbled across an old video from that time… The most fascinating (and slightly unsettling) part of this tech is that, in theory, you could make a facial-recognition camera see a completely different person instead of you. Pretty dangerous stuff! Do you think there’s still demand for devices like this? 🤔

@dontlockmyacc0 @ollama It makes no sense man $100/month and can't even use the best open source model? I'm just gonna bring my $100 to codex or claude instead atleast then I can use their best models without having to pay extra money 💀

@SSSloppyDev @ollama Same here. I even have 3 Max Plan 🤣

DeepSeek-V4-Flash-0731 is now available on Ollama's cloud. This update substantially enhances the model's agentic capabilities: ollama run deepseek-v4-flash:0731-cloud Use it with Claude Code: ollama launch claude --model deepseek-v4-flash:0731-cloud

@OmedVibeCodes You do have to realize that 1. The model does not have vision capabilities so it can not see what 3d it's output looks like. And 2. It's a flash model, any flash model sacrifices capabilities for speed.

DeepSeek V4 Flash being ranked almost equal to Opus 4.8 is actually insane to me. Against Kimi K3, Opus 5, GPT-5.6 Sol, and Qwen 3.8 Max Preview, you give up a LOT by choosing DeepSeek. This is not a 2-point difference in practice. I’ll post the Opus 4.8 comparison next because you need to see this.

@myfolderkarma This guy has clearly never played any sort of tactical shooter game. You don't stand infront of the door waiting you gotta get the peak on them!

@TweeterProMax @Tehjgcfg @QuixiAI @AnthropicAI (2/2) Look who backs Anthropic and OpenAI: Peter Thiel, co-founder of Palantir, the surveillance company that published a manifesto about wanting to take over all state functions basically and make the decisions. That evil motherfucker already wants to take over the world.

@TweeterProMax @Tehjgcfg @QuixiAI @AnthropicAI (1/2) They want to lock down AI and become the sole provider. They want a monopoly on the tech so you can ONLY come to them. They want all your data to better surveil you. Whoever controls the access to high-tier AI in a country holds massive power.

I cancelled my @AnthropicAI accounts. Reason, I cannot support an company that's simply evil.

@QuixiAI @realjohnperi @AnthropicAI Kimi K3 local? Mr. MoneyBags over here with a damn data center in his home 💀

@realjohnperi @AnthropicAI Local of course Can't have my data going to China

@OmedVibeCodes Look up Agent zero by @Agent0ai it's a god tier harness with some insane plugins aswell to make bring it up to a level never seen before 🗿

Things I’d trust with a serious coding task: ✅ Claude Code ✅ Pi Things I’d trust to burn my limits, destroy the output, and somehow blame the model: 🗑️ Hermes Agent 🗑️ OpenCode 🗑️ Qwen Code 🗑️ Codex CLI 🗑️ Kimi Code 🗑️ Grok CLI Same model. Same prompt. Completely different IQ. The harness difference is actually CRIMINAL 😭💀

@philipliao_ Why the heck would you want claude in your team chat. Use a local model for that. You're just giving Anthropic a free entry into all your teams conversations

we installed claude tag and now someone on the team needs to tell it to shut up every 30 minutes

@Ben__Springer @threejs What tools are you using?

Image -> 3D. Massive upgrade! After 37min my agends pipeline combo (opus 5 and kimi k3) builds a comple game asset location pack in detail and convert it in my @threejs player. it build interaction and logic in it. full blueprint incl. material knowledge.

@Arm4x This is amazing. And extremely smart getting all the basic game mechanics down will basically allow you to create any game by just reskinning all the work you've already done. I messed up when I tried as I put a ton of time into how the game looked before how it worked

In my free time, I’ve continued adding features to my Unreal Engine 5.8 AI MCP test. I’m currently using Claude Fable as the orchestrator model, with GPT-5.6 Sol handling subagent execution. New since the last video: - Basic combat system with health, stamina, dodging, parries, executions, and basic enemy AI - Reworked biome generation, with new dense forests, plains, coasts, and mountains featuring more irregular terrain and richer vegetation - Volumetric fog with much lower visibility in the mountains and snow - Improved Nanite trees and grass, now running at 60+ FPS compared to around 30 before

@startracker What engine are you using or by the love of god is this threejs?

Day 08 of building the space game I saw in a dream. 🌙🚀 IT FLIES! The spacecraft I showcased on Day 05 is airworthy, certified and flies under the new realistic physics system I was working day and night for the past few days. ☀️Thanks to astrophysics Professor Fable 5 orchestrating his 4 best PhD Opus 5 colleagues we got: Under the hood: – full 6-DOF rigid-body flight, zero arcade shortcuts – real delta-wing aero: stalls, post-stall drag, the works – fly-by-wire assist so a mouse flies her easily with assist off for the hardcore, oleo landing gear, wheel brakes, speedbrake — you can actually land, an onboard black box records every flight. When physics don't obey, I got the receipts 📼 "Certified" isn't a vibe: 65 automated flight tests green + scripted QC sorties (takeoff, ridge crossing, high-speed decel) flown by the machine before I ever touch the stick. These were also done and orchestrated by the agents under the new skillset that gives them specific controls and camera angles for QC. The other part of this physics system is real weather, wind, lift/drag, shears, turbulence. Yes, I managed to crash several times by underestimating these forces - if high enough, stall recovery was easy, but under 1km, difficult. Skill issue i guess Each part of the spacecraft is programmatic, so you can see the ailerons moving, landing gear retracting, cockpit sliding, even the engine lattices waving. The particle system hasn't been worked on yet, so ignore the random snowflakes going out the engine hahah Another team of agents led by another Fable is currently working on the world generator, so the unfinished island won't be so unfinished soon. Oh, as a bonus, GPT 5.6 Sol built a whole hangar module where we can choose ships and have the tug car haul them to the runway. I'll capture that next time! Next up: a glass cockpit worth flying blind in. Demo video soon. Full avionics!

Let's remove the scammer part and look at the fact that a dev of opencode basically just admitted that they can basically force any model using their service to remotely do things on your PC freely at will and proudly states his ability to do so...

@anumness Are you stupid? It literally says to provide them with a test account or a way to bypass the paywall so that they can review the application. How are they supposed to review your app for approval to be on the playstore if they can't access the full app? This is a Low IQ Post.

Apple allows hard paywalls. No questions asked. Google play however rejected my hard paywalled app because its against their policy.

Running Kimi K3 on my desk? It will require 4 x 512GB M3 Ultra Mac Studios (2TB @ 3.2TB/s). Waiting on active parameter count, but based on the rumours I expect we can run it at 30+ tok/sec with MTP + Tensor Parallelism using RDMA over Thunderbolt 5. Prefill will be slow but 512GB M5 Ultra should make that ~5x faster (expecting in October). It will be on local dot ai with full benchmarks as soon as the weights drop. Comment below for early access - sending access codes out throughout the day.

@alexocheema Sadly I'm too poor for this

@staples46198 @TheAhmadOsman I've never understood people's obsession with needing the newest model. Unless the next model is actual AGI then havibg kimi k3 running locally would be an absolute banger. Heck I use Qwen 3.6 35b and my daily agent running in an agentic harness and it's still amazing + FREE!

@TheAhmadOsman But it will be a 18 month old model by then. No one will care enough. What modelsl are you trying to get running from 18 months ago?

PREDICTION We're gonna get Kimi K3 equivalent intelligence running on a single RTX PRO 6000 in less than 18 months

@OmedVibeCodes @Kimi_Moonshot The secret is to wait for the 27th when they release the open weights and other companies like Ollama add it to their cloud models then buy their coding plan. The limits will be much better

Be careful with the Kimi Code plans — the limits really aren’t as good as I initially thought. Because the pricing is much cheaper and Kimi is a Chinese company, I expected significantly better usage limits. But honestly, that’s not the case at all. Don’t buy it expecting unlimited value or massively better limits. You can burn through them surprisingly fast.

Bye Claude Code Thanks China Thanks @Kimi_Moonshot