Hacker News: bertili

New comment by bertili in "Codex Micro"

bertili — Wed, 15 Jul 2026 16:56:46 +0000

AGI is almost here, but first, one more thing... a keyboard controller!

New comment by bertili in "The infinite scroll may become endangered if controversial Calif. law passes"

bertili — Tue, 14 Jul 2026 07:17:39 +0000

Legislators, please require 10 seconds of load screen with a picture of a tree, for every online video. It worked for cigarette packs.

New comment by bertili in "GLM-5.2 is the new leading open weights model on Artificial Analysis"

bertili — Wed, 17 Jun 2026 10:37:55 +0000

This is GLM 5.2 Max. GLM 5.2 High which use less than half[1] the tokens.

[1] https://z.ai/blog/glm-5.2

New comment by bertili in "RTX 5080 and RTX 3090 Setup: 80 Tok/s on Qwen 3.6 27B Q8"

bertili — Sat, 13 Jun 2026 16:46:07 +0000

Qwen 27b is a compute heavy dense model.

New comment by bertili in "Orthrus-Qwen3: up to 7.8×tokens/forward on Qwen3, identical output distribution"

bertili — Sat, 16 May 2026 07:53:11 +0000

Does this translate into a similar reduction in compute?

What's the catch?

New comment by bertili in "DeepSeek 4 Flash local inference engine for Metal"

bertili — Thu, 07 May 2026 17:58:42 +0000

equals 2 or 3 human brains in power usage. Amazing work!

New comment by bertili in "Qwen3.6-35B-A3B: Agentic coding power, now open to all"

bertili — Thu, 16 Apr 2026 17:53:58 +0000

It's fascinating that a $999 Mac Mini (M4 32GB) with almost similar wattage as a human brain gets us this far.

New comment by bertili in "Qwen3.6-35B-A3B: Agentic coding power, now open to all"

bertili — Thu, 16 Apr 2026 14:34:45 +0000

Is there any source for these claims?

New comment by bertili in "Qwen3.6-35B-A3B: Agentic coding power, now open to all"

bertili — Thu, 16 Apr 2026 13:59:10 +0000

A relief to see the Qwen team still publishing open weights, after the kneecapping [1] and departures of Junyang Lin and others [2]!

[1] https://news.ycombinator.com/item?id=47246746 [2] https://news.ycombinator.com/item?id=47249343

New comment by bertili in "Google releases Gemma 4 open models"

bertili — Thu, 02 Apr 2026 17:40:11 +0000

The timing is interesting as Apple supposedly will distill google models in the upcoming Siri update [1]. So maybe Gemma is a lower bound on what we can expect baked into iPhones.

[1] https://news.ycombinator.com/item?id=47520438

New comment by bertili in "Google releases Gemma 4 open models"

bertili — Thu, 02 Apr 2026 16:49:20 +0000

Qwen: Hold my beer

https://news.ycombinator.com/item?id=47615002

New comment by bertili in "Flash-MoE: Running a 397B Parameter Model on a Laptop"

bertili — Sun, 22 Mar 2026 12:31:46 +0000

Very impressive! I wonder if there is a similar path for Linux using system memory instead of SSD? Hell, maybe even a case for the return of some kind of ROMs of weights?

New comment by bertili in "Qwen3.5: Towards Native Multimodal Agents"

bertili — Mon, 16 Feb 2026 14:21:04 +0000

Better than frontier pelicans as of 2025

New comment by bertili in "Qwen3.5: Towards Native Multimodal Agents"

bertili — Mon, 16 Feb 2026 13:05:16 +0000

Most certainly not, but the Unsloth MLX fits 256GB.

New comment by bertili in "Qwen3.5: Towards Native Multimodal Agents"

bertili — Mon, 16 Feb 2026 12:36:27 +0000

Last Chinese new year we would not have predicted a Sonnet 4.5 level model that runs local and fast on a 2026 M5 Max MacBook Pro, but it's now a real possibility.

New comment by bertili in "GLM-5: Targeting complex systems engineering and long-horizon agentic tasks"

bertili — Wed, 11 Feb 2026 21:08:43 +0000

Exactly. The emperor has no clothes. The largest investments in US tech in history and yet there less than a year of moat. OpenAI or Anthropic will not be able to compete with Chinese server farms and so the US strategy is misplaced investments that will come home to roast.

And we will have Deepseek 4 in a few days...

New comment by bertili in "OpenClaw is what Apple intelligence should have been"

bertili — Thu, 05 Feb 2026 11:27:35 +0000

Surely this is the elephant in the room, but the point here is that Apple as control over its ecosystem, so it may be able to sandbox and make entitlements and transparency good enough, in the apps that the bot can access.

New comment by bertili in "Kimi Released Kimi K2.5, Open-Source Visual SOTA-Agentic Model"

bertili — Tue, 27 Jan 2026 10:26:07 +0000

People are running the previous Kimi K2 on 2 Mac Studios at 21tokens/s or 4 Macs at 30tokens/s. Its still premature, but not a completely crazy proposition for the near future, giving the rate of progress.

New comment by bertili in "Kimi Released Kimi K2.5, Open-Source Visual SOTA-Agentic Model"

bertili — Tue, 27 Jan 2026 10:04:59 +0000

The other realistic setup is $20k, for a small company that needs a private AI for coding or other internal agentic use with two Mac Studios connected over thunderbolt 5 RMDA.

New comment by bertili in "Kimi Released Kimi K2.5, Open-Source Visual SOTA-Agentic Model"

bertili — Tue, 27 Jan 2026 09:02:02 +0000

The "Deepseek moment" is just one year ago today!

Coincidence or not, let's just marvel for a second over this amount of magic/technology that's being given away for free... and how liberating and different this is than OpenAI and others that were closed to "protect us all".