I pulled my kids out of Apogee for similar reasons.
Replying to @BillAckman
My son attended Alpha from 2020 until recently. Most of the families in our original cohort have pulled their kids over the past two years.
Alpha has slick marketing, but there’s been a lot of dysfunction behind the scenes. These are things I personally observed:
1. Persistent teacher turnover.
2. Alpha created a “Dean of Parents” role that acts as a gatekeeper between parents and guides. Direct parent–guide interaction is strongly discouraged.
3. A guide at Alpha’s main campus was disciplined and ultimately fired, in part, for telling a parent that their child was crying daily due to frustration with trying to complete their work in 2 hours. The parent appreciated the transparency. The Dean of Parents objected and attempted to block a scheduled meeting between her and the guide to focus on helping the child.
4. The two-hour academic promise doesn’t hold for most students. Many children struggle to finish their work, some cry regularly, and negative self-talk is common. From what I saw, only about 15–20% of students had an “Epic Day” where all academics were completed. Many kids end up finishing work at home.
5.There’s implicit pressure on kids and parents to speak positively about the school.
Alpha does some things well. They have big aspirations. But right now, the hype doesn’t match the reality. I wish them the best.
1
33
It's insane how much faster @tobi disktree is at gathering my disk usage information compared to the native MacOS Migration Assistant app. Highly recommend.
github.com/tobi/disktree
1
2
42
Can't wait to test M3.1's ability to write prompts for H3.
MiniMax M3.1 Flash: Space Performance + Bunny Speed
🚀🐰 Yes, Space Bunny - which has topped global leaderboards with 200T+ tokens - was an earlier version of M3.1 Flash.
💰Redefining the performance economics of agentic AI. M3.1 Flash achieves frontier-level intelligence across key tasks such as Cowork (Automation-Bench: 62%; ALE: 30%), Coding (DeepSWE: 74%) and Visual Reasoning (BabyVision: 91%), at 1/3 the cost of M3 and 3x the speed.
API pricing (flat rate): $0.15 / $0.60 per 1M input/output tokens, with cache reads at $0.015 / 1M tokens.
Model weights will arrive on Oct 23.
API: platform.minimax.io
M Plan: minimax.io/m-plan
MiniMax Code for desktop: agent.minimax.io/download
34
Using the configuration from the model card (huggingface.co/nvidia/GLM-5.…) results in almost a 2x increase in single-stream tokens/sec compared to the awesome quant by Brandon Music. The setup also resolves the "locklocklock" issue when going concurrent. Many thanks @NVIDIAAI
47
I was surprised how bad it was just walking from the hotel to the Rails World venue. One strung-out guy wanted to fight me for some reason, another was swinging wildly and swearing profusely. Lots of zombie vibes too. Whatever Austin is doing, it's not working.
2017: Austin spent $35 million on homelessness. 2,036 homeless counted.
2025: $118 million. 3,238 homeless.
Spending more than tripled and homelessness went up 59%.
If you service the homeless lifestyle, you will get more of it.
The NGOs that run this system get paid by the problem of homelessness, so they have no reason to end it -- in fact, their only incentive is to make sure the problem grows.
Truth. I bought one after @patrickc posted about it a few years ago. Now it's my favorite gift to give.
After a month of using it, I can attest:
This thing is magic.
Whenever you get bit by some bug like a mosquito or a sandfly, you just put this on the bite for a few seconds and the itchiness goes away and it heals way faster.
79
Wonder which cost more, the company OpenRouter or the domain router.com?
in the end, AI comes down to one number: return on what you spend.
router.com raises it by delivering the right model for every request. our customers are spending 40% less for the same work.
proud of @vral and the @tryramp team, eager to work hard for customers.
87
Day-0 support with official quants in NVFP4 and MLX is clutch. Hope this is the norm going forward.
Aloha! 🌺Introducing Ornith-1.5, a family of open-source LLMs spanning 9B Dense, 35B MoE, and 397B MoE, trained with self-improving strategies.
It achieves state-of-the-art performance among open-source models of comparable size and delivers performance comparable to Claude Opus 4.8 across reasoning, agentic, and coding tasks:
✅Terminal-Bench 2.1 (86.1)
✅SWE-Bench (86 on verified, 65.1 on pro, 79.6 on Multilingual)
✅DeepSWE (56)
✅HLE (44.6)
✅ClawEval (81.4)
✅Tool Decathlon (71.2)
Ornith-1.5 takes a major step toward training foundation models through end-to-end self-improvement, extending the self-scaffolding strategies introduced in Ornith-1.0 into a more complete self-improvement loop: the model proposes new tasks, generates task-specific scaffolds, and produces solution rollouts for reinforcement learning, continuously creating new learning experiences from which it can improve.
All models, along with their quantized versions (FP8, GGUF, MLX, and NVFP4), have been released under the MIT License, enabling unrestricted commercial and research use.
📘Tech Blog: ornith.ai/ornith_1_5.html
🤗Huggingface: huggingface.co/collections/o…
1
49





















