AI Specialist - 🚀 Exploring the frontier of tech & loving every byte! 🐶

Portland
Joined September 2022
Blake Johnson retweeted
the HBM trade could be entering a major shake-up. i asked @tessara_ai what some expert voices were saying on X. here's a quick digest: 1. SK Hynix losing its edge? Reported HBM4 packaging-equipment delays raise execution concerns, potentially opening the door for Samsung to gain share. 2. HBM prices +30–45%? Analysts forecast significant HBM3E/HBM4 price increases in 2027. A major revenue tailwind, although higher manufacturing costs could offset some margin upside. 3. Nvidia's memory appetite keeps growing. Rubin Ultra packs 384GB HBM4E vs. Rubin's 288GB - 33% more memory per accelerator. More HBM demand even without faster GPU shipment growth. 4. China catching up? CXMT is reportedly preparing DDR5 RDIMMs, but KB Securities estimates its manufacturing technology remains 5–7 years behind Samsung. higher HBM prices and content benefit the entire industry. but if Hynix stumbles on HBM4 execution, Samsung and potentially Micron could capture a disproportionate share of the upside. $MU $NVDA
8
7
1
55
10,557
Blake Johnson retweeted
Grok bot, but make it MSN Messenger. buddylist by @safaelmali puts Claude Code, Codex, Gemini and OpenCode in a 2005 buddy list. Each agent gets its own git worktree, you can group-chat them, and pass a half-done task between them. Best part... its FREE. Just bring your own accounts.
14
8
2
22
7,094
Blake Johnson retweeted
DeepSeek’s updated Flash model gained 11.1 points on Terminal-Bench 4.0, rising from 31.2 to 42.3 in the company’s evaluation. Automation-Bench increased from 54.8 to 60.5. DeepSeek reports that the improvements come with shorter outputs and faster task completion. The V4.1-Flash-1017 weights are available under the MIT license.
🚀 DeepSeek-V4.1-Flash-1017 is here! Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API. 🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model. 1/4
4
4
54
2,164
Blake Johnson retweeted
Here is how the future of AI Memory and online learning looks like:
Building AI that learns while it works. We’ve just launched the new MidBrain website. It brings together our research and product, and gives a clearer picture of where we want to take things. For the past year, we’ve been working on a question: How can each new experience make an AI system more capable? When an agent repeats a mistake you’ve already corrected, you end up doing the same work again. We want that correction to change how it handles the next relevant task. The bigger ambition is that the longer a person or team works with an AI system, the more useful it becomes. It should develop capabilities through experience and retain what it has learned as the work changes. At @MidbrainAI , we’re connecting research in continual learning with infrastructure people can use today. Our product turns feedback into learned procedures agents can apply to future tasks. Our research explores how learning can happen during an interaction itself, and how systems can keep adapting without losing useful knowledge. The new site shares our published research, the questions we’re investigating, and the product you can try. If you’re building agents that need to improve over time, we’d love for you to put MidBrain to work. Try it on a real workflow. Challenge our assumptions and tell us where it falls short. There’s a lot we still need to figure out, and working with people using these systems is part of how we’ll get there. Explore the research and try the product: midbrain.ai
2
1
5
989
Blake Johnson retweeted
I was wondering why he didn't drop a Nawwww video in so long lol
Sign to stop Data Centers-
1
1
15
559
Blake Johnson retweeted
3 files that run client outreach for an AI services business: 1. a scorecard. rank niches and businesses by how bad their marketing is. scrape their sites and have your agent compare each one to funnels that work. i have a teardown tool in Boring Funnels that finds the gaps. 2. your offers. how you position and sell, and what's working. 3. a call log. log every call and upload the transcript to your agent. the pitch gets sharper with every call you log.
1
1
10
942
Lol Chamath, don't tell me there was some agenda behind this attack or if he was so clueless. So he really thought CoreWeave will have to pay 15.6% or more on future debt? Who forced them so? What would stop CoreWeave and other neo clouds from NOT taking new contracts at the rates they dont make money?
12
2
64
18,294
Blake Johnson retweeted
It would be a tragedy for embedding models to become narrowly optimized for retrieval in the same way that generative text models have become narrowly optimized for conversational rollout. There is much magic to explore in the realm of representations.
14
25
1
290
13,073
🚨 Doing for Machines what GPS did for Maps: Vangrid GPS did more than show us where we were. It turned location into something software could use. This unlocked navigation, routing, ride sharing, delivery and a lot more. @vangrid_io is going after a similar shift for machines. Not just where is this place? But: ✅ What is there right now? ✅ What changed? ✅ Is the path clear? ✅ Is the entrance open? ✅ Is the data fresh enough to trust? Vangrid is working to make ground data something machines can actually call and use. GPS unlocked location for software. Vangrid could unlock the physical world for machines. That is the bigger picture I can see. Time to find the next missing data set. Dive in.
57
4
70
2,518
Blake Johnson retweeted
Sources: industry execs expect only 7-10 of hundreds of neoclouds to survive, as Nvidia supports them with easier credit terms, site finding, and $20B in leases (@_pheebini / The Information) (Visit Techmeme dot com for the link and full context!)
9
9
2
24
10,257
Blake Johnson retweeted
Replying to @AndrewBolis
Matt Wolfe explains stuff so clean, no cap he's the goat fr
1
17
Blake Johnson retweeted
As global birth rates decline, economist @JesusFerna7026 examines how longer educational pathways, delayed careers, and rising housing costs are making it harder for young adults to start families. He explains how economic policies, particularly restrictive zoning laws, can delay marriage and parenthood by putting homeownership further out of reach. Watch the full conversation with @EconTalker on Hoover+: plus.hoover.org/plus/researc…
Birth rates are plunging around the world, including in countries once known for large families. What's driving the decline? University of Pennsylvania Professor of Economics @JesusFerna7026 joins @EconTalker to examine the mystery of falling fertility and its economic consequences. Watch a new episode of EconTalk from the @HooverInst, @Liberty_Fund, and @Econlib: plus.hoover.org/plus/researc…
2
8
1
21
2,770
Blake Johnson retweeted
Bolt is free for EVERYONE, all weekend long. And you could win $10k! Turn your idea into a product. Submit it by Oct 20 for your shot at the prize. Bolt Builders Week starts now 🏁 Details here 👉 bolt.new/builders-week
114
94
109
589
234,015
Blake Johnson retweeted
you do not need another month to plan your next big idea. you need a weekend to build it. Bolt is free for everyone this weekend, with $10K up for grabs for the best build. What are you making? 👀 Deadline: October 20.
🤝 Paid partnership
Bolt is free for EVERYONE, all weekend long. And you could win $10k! Turn your idea into a product. Submit it by Oct 20 for your shot at the prize. Bolt Builders Week starts now 🏁 Details here 👉 bolt.new/builders-week
9
6
85
23,793
Blake Johnson retweeted
AI optics/networking just had a rough week. in five sessions $AAOI fell ~31%, ciena ~26%, $COHR ~25% some of it is a money worry. at the dbs conference in singapore on oct 12, broadcom and intel both named the cost of capital as the top constraint. then cpi came in hot, 3.7% vs 3.6% expected. plus some profit taking but look at what taiwan's server makers filed for september, vs a year ago: wiwynn +100% wistron +87% quanta +77% wiwynn basically doubled, NT$172.6B vs NT$86.3B. revenue lands when the racks ship and those orders were placed months back, so this tells you the money was still flowing through september our gpu rental index across 6 providers is up ~17% in a month, which supports my read that so far this is a scare about financing, and the AI demand numbers haven't turned i'm watching oct 28, when microsoft guides capex, with alphabet and amazon the same week. if those guides hold or rise.
11
2
30
5,603
Blake Johnson retweeted
old pitch: "we'll build you a chatbot". new pitch: "every missed call gets a reply in under a minute". old: "i'll build you a new website". new: "i'll build a funnel that makes your slow months full". old: "an AI marketing evaluation". new: "i'll find where you can get more booked jobs from the traffic you already have, with no more ad spend". you find the chatbot interesting. the owner finds their slow months interesting. say what it does for them, in their words.
5
4
1
83
4,942
Blake Johnson retweeted
Meta has released the weights for Muse Spark Open, a model built for coding and multi-step agent workflows. It has 816B total parameters, with 42B active per token, and supports up to a 1M-token context window. It accepts text, images, video, and audio, with text output. Weights are available on Hugging Face. Hosted access is available through providers including Fireworks AI, OpenRouter, and Together AI. Developers can customize and deploy the model under Meta’s Muse Model License.
Today, we’re releasing the weights for Muse Spark Open, giving developers more control over how they run, customize and deploy the model. (1/5)
3
4
2
20
1,541
This guy literally built a 24/7 AI workforce on Hermes Agent and runs it from one screen called an Agent OS. Most people install Hermes and quit fast. Not because it's weak. Because they can't see what it's doing. Here's the workforce: → Each bot gets its own cloud computer and works while you do something else → Apollo, a voice agent you can interrupt and get a daily briefing from → Oracle pulls fresh news every 24 hours and turns it into posts → Muse reads your view counts and suggests what to make next → Astros watches competitors and hands you fresh angles → Lead finder. Describe your ideal contact, it pulls the list via Hunter, drafts the email, you approve. → Sub-agents. One researches while another drafts, side by side. The plugin that matters: Claude Subscription Direct SDK. Run Opus 5.5 on your Claude Pro or Max plan. No separate API key. Hermes v0.20 raised tool steps per task from 90 to 500. It keeps going way longer now. The memory trick: it only saves useful stuff to Obsidian. So it never fills with junk. How he built every piece: "Hey Hermes, I've got this idea. Set it up in my Agent OS." Start with one problem that eats your time. Build that. Then add the next. Want the SOP? DM me. 💬
1
1
1
1,457
Blake Johnson retweeted
I just built a free tool that helps Claude and other agents watch videos. It's called Eyes. Give it a video file or link. It processes the video once and gives your agent: → Sheets of labelled frames to look through → Every spoken word, with a timestamp → Text shown on screen, with when and where it appears → Details about the sound and editing style Claude can't actually watch videos. When it tries, it may only check a few frames and guess what happens between them. Follow-up questions can make it start over, using more time and tokens. It may also miss information that is shown but not spoken, such as commands typed in a coding tutorial. Eyes does the work once. Your agent can then search the video, find the right moment and zoom in to read small text or code. Runs locally. Free and open source.
12
5
2
71
10,491
Blake Johnson retweeted
GitHub has officially gone insane. You can now build an entire AI content factory without hiring a single person. 10 ridiculously powerful repos: 1. OpenCreator Give Codex a creator workspace to write scripts, generate images, create voiceovers, translate videos, and edit content. github.com/krillinai/OpenCre… 2. Social Media Research Skills Give Claude Code and Codex skills to find standout posts, study competitors, mine comments, and pull hooks from video transcripts. Powered by ScrapeCreators. github.com/ScrapeCreators/so… 3. Local Deep Research Turn a topic into a research report by searching the web and your documents. Works with local and cloud AI models. github.com/LearningCircuit/l… 4. FLUX.2 Generate and edit images for thumbnails, social posts, and campaigns. The Klein 4B models come with Apache-licensed weights. github.com/black-forest-labs… 5. Kokoro Turn your scripts into AI voiceovers using a compact, open-weight speech model. Build narration directly into your content pipeline. github.com/hexgrad/kokoro 6. ACE-Step 1.5 Generate music from text descriptions, create instrumental tracks, and regenerate sections of audio. Runs locally on supported hardware. github.com/ace-step/ACE-Step… 7. Short Video Factory Turn prompts and your footage into batches of short videos, with AI-written copy, voiceovers, automatic editing, and subtitles. github.com/YILS-LIN/short-vi… 8. AI YouTube Shorts Generator Turn long YouTube videos into vertical Shorts with AI-picked highlights, transcription, and automatic cropping. github.com/Anil-matcha/AI-Yo… 9. Podcastfy Turn source material into AI podcast conversations. Customize the dialogue and generate audio in multiple languages. github.com/souzatharsis/podc… 10. Social Media Agent Turn a URL into X and LinkedIn posts. Review the drafts, make changes, and schedule them through an AI agent workflow. github.com/langchain-ai/soci… Which part of your content workflow are you automating first?
49
437
8
2,716
145,860