ABE DIAZ
@abe238
Followers
3K
Following
5K
Media
808
Statuses
17K
Disaster Relief. Tweets are my own and do not indicate opinion, also RTs are not endorsements. 🇵🇷
Seattle, WA, USA
Joined November 2007
😁
➡️ Abe Diaz from Amazon’s disaster relief team discusses how we're innovating ways to respond to communities after disaster strikes.
4
4
35
This is gonna save folks lots of effort scraping HN. ALS smart of HN to move all that scraping traffic elsewhere. Brilliant!
0
0
0
Brilliant!
stop doing demos. do reverse demos. "open the app and let's do this together" I've had 50+ conversations like these in the past 2 months: I tell them exactly which buttons to click. they tell me exactly what's not clicking. best part is zoe reads the transcript after and
0
0
0
Dr. @AndrewYNg One thing that's been frustrating: the Gemini Deep Research API (Interactions endpoint) is completely separate from the standard GenAI SDK, and coding agents constantly get it wrong. I submitted a PR adding docs for it: https://t.co/guSiLoULEa If it adds value,
github.com
What Adds documentation for Google's Gemini Deep Research agent via the Interactions API — a separate REST endpoint from the standard generateContent API that runs an autonomous multi-step ...
0
0
2
Yeah PRDs are more alive than ever. But you write them by: - Chatting with AI - Voice dictating stream of conscious thoughts - Pasting in reference materials Not by typing out letters one by one in your keyboard.
PRDs are more alive than ever. When the cost of implementation goes down, describing what to implement and why is where all the leverage goes.
25
18
187
Guys @karpathy is gonna tinker with claws this weekend. I can’t wait what he publishes by next Wednesday. VibeClaw incoming! 😂🙏
Bought a new Mac mini to properly tinker with claws over the weekend. The apple store person told me they are selling like hotcakes and everyone is confused :) I'm definitely a bit sus'd to run OpenClaw specifically - giving my private data/keys to 400K lines of vibe coded
0
0
1
Hearing of a new benchmark the “LennyTest” as @bcherny said the @claudeai team used the @lennysan post of 50 non-technical use cases for Claude Code as a benchmark for testing Cowork before release. Brilliant. Ps: @AcquiredFM mention too!
2
0
4
If we follow this trend of almost a new model every week from the top labs; that means at least 52 weeks of iterations in 2026. The speed of innovation has never been the same.
This is Claude Sonnet 4.6: our most capable Sonnet model yet. It’s a full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design. It also features a 1M token context window in beta.
0
0
1
Holy ……
I'm joining @OpenAI to bring agents to everyone. @OpenClaw is becoming a foundation: open, independent, and just getting started.🦞 https://t.co/XOc7X4jOxq
0
0
2
Yup. Totally get that.
Short musings on "cognitive debt" - I'm seeing this in my own work, where excessive unreviewed AI-generated code leads me to lose a firm mental model of what I've built, which then makes it harder to confidently make future decisions
0
0
0
Wise advice 👇👇👇
If you want to get hired as an AI Product Manager in 2026, do this one thing: Go build something. I just visited Harvard Business School to lead a session on PM resumes. We reviewed 30+ resumes and had a conversation about what "AI Product Manager" actually means today (hint: it
0
0
0
And this is the truth! 🇵🇷❤️🙏
BREAKING: At the Super Bowl halftime show, Bad Bunny says he prefers Opus 4.6 for vibe coding but turns to 5.3 Codex for gnarly engineering problems. “Opus pa’ vibear, Codex pa’ lo heavy” - he said
0
0
1
This is insane. When does a major tech company drop a 2.5x faster product on a Saturday at 10:45am? The pace of shipping is ridiculous. @_catwu @bcherny @alexalbert__ @claudeai 🤯
Our teams have been building with a 2.5x-faster version of Claude Opus 4.6. We’re now making it available as an early experiment via Claude Code and our API.
0
0
0
Pete @steipete you made it to new model release notes. 🙏”Supports OoenClaw”.
🚀 Introducing Qwen3-Coder-Next, an open-weight LM built for coding agents & local development. What’s new: 🤖 Scaling agentic training: 800K verifiable tasks + executable envs 📈 Efficiency–Performance Tradeoff: achieves strong results on SWE-Bench Pro with 80B total params and
0
0
0
This paper makes AI agents that use computers much harder to hijack by forcing them to plan before seeing screens. The big deal is the agent’s allowed steps cannot be rewritten by a webpage, which is the main way prompt injection wins. Because these agents read whatever is on
3
17
94
There’s a weird narrative floating around that using AI for simple tasks makes you 'incapable' or 'lazy.' I think we’re confusing Capability with Capacity. Yes, I can read a map. Yes, I can cross-reference three emails. But when I’m balancing a last-minute trip, chasing three
24
6
80
Qwen now powers DINQ — an AI-native professional network helping AI professionals build trusted profiles and connect with opportunities more efficiently.
You’ve done real work. But most of it is hard to see. DINQ brings your projects, code, and research onto one card. No self-promotion. Just real signals. Build your DINQ → https://t.co/IaEtN0Vkab
#DINQ
9
34
393
Introducing MedGemma 1.5, an open-access model for multimodal medical use cases It expands the tasks and data formats it can understand (high-dimensional medical imaging, EHRs, anatomical localization with bounding boxes, etc) https://t.co/m2wjry7M8s
28
132
1K