Sabitlenmiş Tweet
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc
18.2K posts
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc
@matt_boh
Cybersecurity Engineering | Arch btw | prev CEO of Nootropix | Biohacking | 🦐🦾 | e/acc | #AI | 21e8
San Diego Katılım Ekim 2008
7.5K Takip Edilen3.6K Takipçiler
Hermes Agent is a piece of shit code like the rest . I have no clue what people are praising these shit code bases for .. they are the most fragile pieces of crap vibe coded nonsense. They fall apart so easily and quickly on the first hint of complexity ..
A fucking do while loop outperforms them every time.
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
anyone thinking about, learning, or already working with agentic systems, you should know this.
the first few steps of your setup matter more than any model or framework you pick later. get them right and you never lose your flow.
the foundation nobody posts about:
> 1. tailscale. a private mesh network across every machine you own. laptop, desktop, rented node, all on one secure tailnet, reachable from anywhere. nothing else works well until this does.
> 2. termius, over that tailnet. one SSH client that reaches every node, phone included. you are never away from your stack.
> 3. tmux. persistent sessions. disconnect, close the laptop, come back, every session exactly where you left it. agentic work runs long, your terminal has to survive that.
> 4. a private git repo. the one i am most glad i found. it is the memory layer across all my agents, they pull, they work, they merge back, the codebase stays alive between sessions. context that would die in a chat window lives in the repo instead.
> 5. script everything from day one. ssh aliases for every node, setup scripts, the boring boilerplate automated. if you will do a thing more than twice, it is a script.
everything past these five is decorative. know these cold.
and the habit that ties it together: ask the AI itself. for the config, for the error, for any of it, let the agent do the lifting, then double check what it hands you.
lock the five, build the habit, and you make it. skip it, anon, and you ngmi.
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
GITHUB JUST CREATED AN OFFICIAL CERTIFICATION FOR THE MOST IN-DEMAND DEVELOPER ROLE OF 2026.
It is called Agentic AI Developer.
GH-600.
And it is the first formal signal that running AI agent teams is now a recognized engineering discipline with a credential behind it.
Not a prompt engineer.
Not a vibe coder.
An Agentic AI Developer.
The person who operates, supervises, and integrates AI agents across the entire software development lifecycle.
The person who knows where agents fail in production.
The person who understands how to build autonomous workflows that do not introduce catastrophic failure modes into CI/CD pipelines.
The person every engineering team is going to need and almost none of them have right now.
GitHub certifying this role changes the hiring conversation permanently.
Before GH-600: "Do you work with AI agents?" is an interview question with no standard answer.
After GH-600: the credential tells the hiring manager exactly what you know and what you can do before the interview starts.
The engineers who get certified in the first wave of GH-600 will have a credential for a role that has more demand than supply for the next 3 to 5 years.
The engineers who wait until it is mainstream will be competing with everyone who moved first.
If you are already working with GitHub Copilot or building agent-driven workflows you are already doing this job.
GH-600 is how you prove it.
Bookmark this.
Follow @cyrilXBT for every AI certification worth your time the moment it drops.
Microsoft Learn@MicrosoftLearn
We’re introducing a new GitHub Certified: Agentic AI Developer (GH-600). As AI agents become part of modern development workflows, this role-based certification focuses on how developers and teams operate, supervise, and integrate agents across the SDLC. If you’re already working with tools like GitHub Copilot or exploring agent-driven workflows, we’d love your input. Learn more and get involved. msft.it/6013vRHHZ
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
@XFreeze it seemed like a while back I couldn’t use it with hermes but I saw something about hermes integration today so maybe…
English
xAI just dropped a massive new offer for SuperGrok Heavy plan
67% off for 6 months - now $99/month instead of $300/month
Includes:
• Near-unlimited usage
• 16x AI agents in Expert Mode
• Early access to new features
• Dedicated support
• Access to Grok Build Early Beta
• xAI’s most powerful reasoning tools
And this is just the beginning
There are a lot of new Grok models and features coming very soon
xAI is moving insanely fast right now
Subscribe while the offer is still available
English
@Teknium oh snap! big grok incoming!! i know grok 5 is gonna rip too
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
Give our early preview of Computer Use (with ANY model) a try today!
Built into the latest Hermes Agent and powered by @trycua - opens the door to any model, not just the frontier models in special modes - to control your actual computer.
Best part, it doesnt take over your PC - you can continue to work and operate with full control of your keyboard, mouse, and screen - works entirely in the background!
Nous Research@NousResearch
Computer use with any model Hermes Agent × @trycua
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
The honest secret of running AI agents inside a real company is this: the model is not the bottleneck anymore. The bottleneck is what happens when the agent is wrong.
I run agents across several of my companies. They sort emails, manage dashboards, block bots on X, draft replies, summarize calls. The first version is always magical. The tenth version is where you learn the real lessons.
The model is rarely the problem. The problem is that nothing in the stack tells you, in production, that the agent quietly drifted. It does not crash. It does not error. It just becomes slowly worse at the job, and three weeks later you realize half of its outputs are subtly wrong.
What you actually need is unglamorous: evals you trust, logs you can search, the ability to roll a single agent back to last week, and a human review queue for anything that touches money, legal text or a customer. Most teams skip all four because they are not as fun as a new model.
The companies that win with agents will not be the ones with the smartest model. They will be the ones whose engineers treat agents like junior employees with bad memory and worse judgment, and build the supervision around them accordingly.
Intelligence is cheap now. Accountability is what will be priced.
English
Elon Musk after removing Anthropic's biggest bottleneck while simultaneously killing their #1 competitor (OpenAI).
Claude@claudeai
We’ve agreed to a partnership with @SpaceX that will substantially increase our compute capacity. This, along with our other recent compute deals, means that we’ve been able to increase our usage limits for Claude Code and the Claude API.
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
New model from Jackrong on @huggingface
Excited about this one! @KyleHessling1 and Jackrong are back with Qwen3.6 35B A3B distilled on Claude Opus reasoning 🔥
Model size is 71.9GB
GGUFs should be released very soon 👀
huggingface.co/Jackrong/Qwopu…
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
Introducing Mirage, a unified virtual filesystem for AI agents!
6 weeks. 1.1M+ lines of code. We rewrote bash from the ground up so cat, grep, head, and pipes work across heterogeneous services. S3, Google Drive, Slack, Gmail, GitHub, Linear, Notion, Postgres, MongoDB, SSH, and more, all mounted side-by-side as one filesystem.
Bash that AI agents already know works on every format! cat, grep, head, and wc parse .parquet, .csv, .json, .h5, even .wav! One pipe can stitch S3, Drive, GitHub, Slack, and Linear together, same Unix semantics throughout.
Workspaces are versioned too. Snapshot, clone, and roll back the whole thing with one API call. A two-layer cache turns repeated reads into local lookups, so agent loops stay fast and cheap.
Drop a Workspace into FastAPI, Express, or a browser app. Wire it into OpenAI Agents SDK, Vercel AI SDK, LangChain, Mastra, or Pi. Run it alongside Claude Code and Codex.
Site: strukto.ai/mirage
GitHub: github.com/strukto-ai/mir…
#AIAgents #OpenSource #AgenticAI #Strukto #Filesystem #VFS
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
Something's totally off about the number of data centers being built (over 3,000 right now) and the sheer size and compute power they represent.
They are massively OVER-building capacity that can't possibly be met by customer demand for compute.
And customer revenues can't possibly recover the financial investment needed on these projects.
There's clearly some other plan afoot, and I don't yet know what it is. It involves massive compute, but not merely to serve inference or hosting databases and corporate data. There's a much larger plan at work here.
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
@HealthRanger This is a market share race for the compute layer.
Hyperscalers are trying to lock up future AI capacity before the bottlenecks harden: power, land, chips, fiber, transformers, and grid access.
That physical constraint is central to The Faster Curve:
x.com/cogan/status/2…
James Cogan@cogan
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
Hermes Vault 0.6.0 is out.
And this one is nasty.
I added OAuth PKCE login + token auto-refresh.
So now Hermes Vault can:
🔹open a browser login flow
🔹store access + refresh tokens automatically
🔹refresh near-expired tokens before they break
🔹support Google, GitHub, OpenAI, or custom providers
🔹expose OAuth login/refresh as MCP tools
🔹audit every event without leaking secrets
This is the difference between:
“my agent has a pile of API keys”
and
“my agent has a credential system.”
That matters.
A lot.
Repo: github.com/asimons81/herm…
English
@KyleHessling1 I like uncensored models ! I will check this out
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
Hello again, everyone!
Our latest Qwopus3.6-35B-A3B-v1 is now live, and it is once again breathtaking!
Full HF space benchmark showcase and write-up is in the comments, so you can make conclusions for yourself! GGUF Link below!
We had some issues to fix in our training with this one, as we hit it much harder with more sophisticated processes. Which is why it took so long to complete, but it was totally worth it! This model performs like a significantly larger model, and runs at an incredible 162tps on a single 5090 straight out of the box!
The web front ends are exceeding the capability of the base Qwen 3.6 27B in my opinion. This model outputs significantly more complete web pages with the exact same prompts in one go. It is more verbose, but in a good way for this sort of work!
Thinking efficiency seems to be improved over the base 35b as well, and at no cost to output quality or depth, which is incredible! It thinks more carefully when the prompt rewards depth, but the output quality reflects this rather than waiting an eternity for a basic answer, like with the base models!
Those with less VRAM can run this at very usable speeds, partially offloaded to system RAM, and it performs comparably if not better than the base 27B dense version. So if you can't fit the 27B on your GPU, this one offers enhanced capability and speed when offloaded. Or if you can fit the 27B, this one seems to perform better at over 3x the speed!
Give
I truly believe Qwopus3.6-35B-A3B-v1 at Q5_K_M is one of the strongest one-shot front-end and reasoning models you can run on consumer hardware right now! Would love to see what you all can make with it!
Really excited to hear your feedback! GIve it a thorough benchmark and let us know any idiosyncrasies you may discover!
I am all ears in the comments, and new 27B is still training but will be complete soon! huggingface.co/Jackrong/Qwopu…
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
Derek Mobley applied to over 100 jobs. He was rejected from every single one. Several rejections came at 1am, within minutes of submitting.
He just became the lead plaintiff in the largest AI lawsuit ever certified.
May 2025, Judge Rita Lin granted preliminary certification of a nationwide ADEA collective in Mobley v. Workday. Workday's own court filings represent that 1.1 billion job applications were rejected through its software in the relevant period. The court discussed potential class size in the hundreds of millions.
If you're over 40 and you applied to a Fortune 500 in the last 7 years, your application was probably processed by Workday. You may be in the class.
The legal precedent matters more than the headline number. For decades, the vendor screening applicants for an employer was not directly liable under Title VII. The employer was the only defendant. In July 2024, Judge Lin ruled the AI vendor itself qualifies as an "agent" of the employer and can be sued directly. First time. The "we're just the tools" defense evaporated in a single ruling.
Same precedent now extends to every HR tech AI vendor in the pipeline. Greenhouse. Eightfold. HireVue. Paradox. None of it is priced into any of their valuations.
Combine that with the rest of 2024. Air Canada lost in February for $812 because its chatbot hallucinated a refund policy, killing the chatbot-as-separate-entity defense. iTutorGroup paid $365K to the EEOC, confirming the algorithm doing the discriminating moves liability nowhere. Gemini cost Alphabet roughly $90B in market cap in days for one weekend of bad image generation.
Every legal shield around AI in production got tested in court and lost. The AI PMs interviewing for foundation model roles can recite all four by month. Most engineers shipping AI at work cannot.
Aakash Gupta@aakashgupta
Reports are that Anthropic has a dedicated safety & ethics interview round. Here's how to build the first principles to crack it. Start with the distinction most candidates miss. AI safety is stopping the model from causing harm. Mechanical and observable. A guardrail that blocks medical misinformation. A confirmation gate before an agent sends an email. AI ethics is deciding what the model should and shouldn't do. Upstream and often invisible. A policy that the model won't draft termination letters. iTutorGroup's hiring AI worked exactly as designed. The ethics of the design were the failure. Then learn the sizing framework. The candidates who scored highest in our Cohort 3 mocks all used SHIR in the first 90 seconds of every safety question, before they proposed anything. S: Severity. Physical harm sits above discrimination sits above embarrassment. H: Harm scope. 10 users versus 10M users is a different response. I: Immediacy. Active harm or latent risk. Sets response speed. R: Reversibility. Can the action be undone. Determines whether you ship with monitoring or add hard confirmation gates. Pause, write the four letters on a notepad, then come back with structure. That move scored higher than any other in mocks. Then memorize four precedents. Cite by name and month. Air Canada chatbot, Feb 2024. British Columbia tribunal held the airline liable for a chatbot hallucinating a bereavement fare. The defense that the chatbot was a separate legal entity got rejected. Companies own their AI's representations. iTutorGroup, Aug 2023. $365K EEOC settlement. Hiring AI auto-rejected women 55+ and men 60+. About 200 qualified applicants screened out. Bias liability lands on the employer even when the algorithm does the discriminating. Mobley v. Workday, July 2024. First AI vendor held directly liable as an "agent" under Title VII. Collective certification followed under ADEA in May 2025. Vendor liability is no longer theoretical. Gemini image gen, Feb 2024. Alphabet shed roughly $90B in market cap in the days after the pause. Sundar called the outputs "unacceptable." The cost of acting is almost always lower than the cost of being seen as not acting. Then practice the three moves that consistently scored highest in our cohort mocks. Tier the response. Three options side by side with a cost on each. Binary "pull or ship" reads junior. Reframe under pushback. When the VP says "wait until earnings," the highest-scoring answer reframed revenue to headline. End with documentation. If leadership overrides you, write a memo to the manager, the safety lead, and legal. On the record. The full kit: → Video version on YouTube: youtube.com/watch?v=RaBw5S… → Coaching with Ankit Virmani (AI PM, Uber), Prasad Reddy (ex-CPO), and Dr. Bart Jaworski (12,000+ PMs coached) in Cohort 3, which opens tomorrow: landpmjob.com → Mock walkthrough and writeup on the newsletter: news.aakashg.com/p/safety-ethic…
English
🇨🇾🇵🇭🇪🇷🇵🇺🇳🇰| MattGPT | e/acc retweetledi
Herm: All-in-one TUI built for Hermes. Chat with your agent and watch it manage everything it touches.
11 tabs. 42 themes.
My submission for the creative hackathon.
npm i -g herm-tui && herm
github.com/liftaris/herm
@NousResearch @Teknium @imbabybrooklyn
English
@mr_r0b0t @NousResearch yo figured you may be interested in my complete feature and function parity Rust rewrite of Hermes
GitHub.com/sheawinkler/he…
English
For all the people who were curious about running 96+ deepseek-v4 powered @NousResearch Hermes agents in parallel, I packaged that skill up and have open sourced it on GitHub.
Hope you enjoy DeepSwarm!
English