The future may belong not to those who know the most, but to those whose identities can survive being transformed by what they learn.

I appreciate @TheAhmadOsman. Respect for the shout out. Your reason I got my GPUs before I was solely a Mac Guy. Keep pushing. I like how you support and encourage the local AI community. Appreciate the positivity. I will also take the time to shoutout the community. I would be nothing without them. @mr_r0b0t @Blackwellboy @NeoAIForecast @Tech2Wild @MichaelGannotti @aijoey @BosonJoe @with_gene2626 @volatilemarkts @ivanfioravanti @thatcofffeeguy @WescheNex1q @SpaceTimeViking @yume_arasaki @0xSero @GordianKnotDev @LytninCrypto @light_foundry @MichaelHutu @thedatabunny @StefanMaier @anemll @Q_Beaux @SixZzshOtRipZz @anuntrapid_auto @GumbiiDigital @GBminA @ayayalar @chrittazz @jonathan_conn @HundtRichard @Dave_Charland @RyanFogleOP12 @MarkSunner @haydonryan @bitk0rns @Defenestrator22 @Jr1Mj5zqz8yqy @kkiran @NsitnovSitnov @DJLougen @imost @wikiwayne @CosmicRaisins @ViC305 @ashxhart

Thank you Mr Boss Man @TheAhmadOsman

Coming soon: GLM 5.3 Flash EXL3 update 🔥

It’s funny I don’t even know where to spend $20k on hardware right now??

Thank you, we pushing

tl;dr - the new Mac is VERY fast across the board for low concurrency, but scales horribly to higher concurrency on both models at this point. Making great headway on the Mac so far, have had clankers on it since the moment I turned it on, benchmarked all inference engines and quants of Qwen 3.8 Flash and GLM 5.3 Flash for a solid baseline and am having them work on improved kernels for both models, already seeing massive improvements. Initial findings confirm oMLX is the best engine right now for this, which is what I’m basing my optimization work around at this point.

I wonder what’s the artificial Analysis score @ArtificialAnlys for Mimo v2.6 Flash

Space Bunny Alpha First Test – What IS This NEW Stealth Model? piped.video/gRnRCv1t0r8?si=E3I2… via @YouTube

AI painting of the day. Created with ChatGPT Image 2.5

opus 5.5 rebuilt the last of us main menu i could barely tell them apart it handled lighting way better than astra did for me. it's also cheaper

Disturbingly awesome

GLM 5.3 Flash rendention of Dubai Voxel. It can do alot better its a start

I saw a comment recently that said ‘harness is everything’, and it couldn’t have been more accurate when it comes to models with multiple thinking modes. Using Bonsai 2 in my example since it encapsulates this point so well. Spread on coding bench scores are simply out of this world depending on so many factors such as your app, harness, thinking mode, token cap, all of which I’ve tested to get to the bottom of this. Bonsai 2 does code at 27B level, on medium with minimum 8k+ cap which is not the default setting for most app/harness that was hurting performance badly. 27B have similar issues to be fair. Generally xhigh only performs best when given outrageously high cap which nobody will sit through, so if you are getting bad results with thinking mode on, try higher cap and different modes to see if it helps your use cases.

Qwen just nails local AI architecture. There’s something they understand about building models for real-world local compute that most other providers are still missing.

If you put $10,000 into tech hardware in January 2026, here is what that money looks like today: > DDR5 RAM: $70,000 > Mac Studio M3 Ultra (512GB): $30,000 > 3x DGX Spark: $24,000 > 1x RTX 6000 Pro: $18,000 > 2x MacBook Pro M5 Max (128GB): $16,000 Memory was the real trade all along.

deterministic reproduction is not a promise that every user can trigger the bug it is a controlled claim: known start state, environment, inputs, and ordered actions let another person test the same failure guide: samelogic.com/blog/determini…

Jev but Opensource