Devin Prater

This is how you shut disabled people out. github.com/anthropics/cla… Blind person reports issue. No one cares. Bot closes and locks issue. To be clear, it's not just this one company or issue. #disability #blind #foss #tech #anthropic #claude #code #claudeCode @AnthropicAI

[Blog Post] Apple Introduces AirPods Max 2 with H2 Chip and New HDR Amplifier dlvr.it/TRWvJN

I know there is some overlap between open source and anti-AI activists, but I have a hard time reconciling it. My million+ open source LOC were always intended as a gift to the world. Yes, I would make arguments about how it would strengthen our communities, and the GPL would prevent outright exploitation by our competitors, but those were to allay fears of my partners to allow me to make the gift. AI training on the code magnifies the value of the gift. I am enthusiastic about it! Some people do look at open source as a tool for social change, career advancement, or reputation building, but those are all downstream of the gift.

This weekend was something special. We ran a 16 hour workshop teaching people how to use Git and GitHub accessibly. From the very beginning. What a repo is. How to clone, commit, push, and collaborate. All of it done with a screen reader. Then we went further. Students learned to build their own custom agents with VS Code and GitHub Copilot. @JeffBishop built image agents that are already in the workshop repo and ready to use. People came in not knowing Git. They left feeling confident enough to contribute to open source. Thank you to @JeffBishop, @mikedoise, and every single person who made this happen. This is what accessibility education looks like when the community shows up. #a11y #GitHub

I miss the days when I, an average US citizen, could go literal weeks without hearing or even thinking about the United States president.

Kelly Wilkinson was being abused by her husband. She tried to go to the police. They told her to “cool off”. To give her husband a “break”. They accused her of cop shopping. Her husband stabbed her over 20 times and set her on fire. The cops were complicit in her murder:

Having a local LLM gives me the peace of mind to do things like this. Using gptel-agent with Ollama running qwen3.5:27b: So I was curious if it can do a simple speed test within Emacs. I then wanted the results in the same metrics that browsers use when downloading files, and it checks out. The model can call tools using gptel-agent, which is usually running Emacs Lisp, but it can also run bash. I wouldn't put this on making anything complex, but for simple web searching and doing fun things like this, it's surprisingly good. Speed Test Results: 58 Mbps (7.2 MiB/s) Download times: • Web page (5MB): 0.7s • Album (50MB): 7s • HD movie (1GB): 2.3 min • OS image (4GB): 9 min • Large game (50GB): 1.4 hours #Emacs #gptel #gptel-agent #foss #ai

@GeanKassidy @chatgpt21 I'm fairly sure. They wouldn't hold out two models for just Codex.

@chatgpt21 Will more than codex users get it though?

Omgoodness it’s true 🥹 GPT 5.4

Why does everyone force introverts to "leave their comfort zone" but no one forces extroverts to shut up for awhile?

To everyone who has ever asked, or even thought, “Why didn’t she go to the police? Why didn’t she report?” Sincerely, from the bottom of my heart, fuck you.

HOLY MOLY I just tested Qwen 3.5 9b and it's the future 8GB of RAM. Less than Chrome

@simonw I run agents through Telegram daily. Screen readers struggle with streaming because ARIA live regions get overwhelmed. Better to buffer responses and update in chunks with clear state indicators.

I'm very excited that we've released version 1 of Perspective Server for Mac. This creates an Ollama like experience for running Foundation MOdels in other applications. It even works with OpenWebUI. github.com/Techopolis/Per…

Starting internal testing on Perspective Intelligence 1.6 with user and tool agents. This is exciting!

The Qwen 3.5 small model series is now available ollama run qwen3.5:9b ollama run qwen3.5:4b ollama run qwen3.5:2b ollama run qwen3.5:0.8b All models support native tool calling, thinking, and multimodal capabilities in Ollama.

Condoms should be free. Tampons should be free. And since we telling the truth So should pads Soap Toilet paper Birth control Pregnancy tests Diapers Baby formula Narcan EpiPens Hearing aid batteries Disability access tools Basic hygiene and survival should not be a luxury. If you gotta bleed, breathe, eat, hear, or stay alive it should not come with a price tag. We got money for wars, bailouts, and billionaires but not for people to exist with dignity? Make it make sense. Free the basics.

@Eduardopto @simonw I agree, I hate streaming so I turn it off where I can.

@simonw Streaming chat accessibility is genuinely unsolved. ARIA live regions help but fast token streams are brutal for screen readers. Curious what you find, most teams just ship and ignore it.

Surprise new feature in Telegram, inspired OpenClaw et al using Telegram as a UI Anyone know how well this work with screen readers? I'm still not confident I know how to make streaming chat UIs like this available in an accessible way

In case you’re wondering just how messed up Windows is when it comes to running AI inference, check this out: Qwen3.5-112B A10B running on a Windows box via LM Studio generates about 20 tokens / second. The EXACT SAME HARDWARE, just running Linux with LM Studio, generates 70 tokens / second. Same GPU, same RAM, same CPU, same box. Literally the same box. Two totally different results. That’s how bad Windows is. It even dumbs down AI.

@xiaoyiz_uw Thanks for letting me know. I think I did send feedback through there about the Android issues at one point.

@devinprater In the future, you may also use Settings > Send Feedback to report any accessibility issue. Sorry again for the experience!

You will enjoy it ;)

@xiaoyiz_uw Thank you so much. I mainly use Android, but for iOS there is [lots of documentation](developer.apple.com/design/human-i…) on accessibility for developers, and [Google has docs too](developer.android.com/guide/topics/u…). A few people have even made this Claude Code extension: github.com/Community-Acce…

@devinprater @devinprater thank you for the feedback! Sorry for the experience. I can reproduce that on iOS when thinking is off. The focus auto-moves to "copy" button; we have to swipe several times to hear the reply message due to the navigation order. I shared this with our team for a fix.