use alchemy.run for anything infrastructure no matter the provider
even cloudflare knows alchemy is the superior way for agents to use cloudflare
> Fable 5.2 high with /grill-me, then /to-spec and /to-tickets
> Opus 5.5 medium subagents with /implement
> Cloudflare CLI for hosting & database
IYKYK
1
11
938
Replying to @thefubhy
you can replace wrangler w/ cf cli today, or if you're like me, replace it w/ alchemy today
145
confession: 90% of my usage the last 2 months have been deepseek and i only use claude in specific scenarios
deepseek v4.1 flash and claude opus/fable is the only 3 models you need.
I think we're at a point right now where China's open weight models are actually catching up AND surpassing western models
2
8
720
ecom niggas will look at this and still try to convince you that AI can't write good copy
‼️ BREAKING: Oracle Issues Emergency Java Update for Critical 9.1 HTTPS Trust Vulnerability
Details: cybersecuritynews.com/emerge…
Oracle has released an emergency Java security update to address CVE-2026-91403, a critical vulnerability affecting how Java applications trust digital certificates. Published on October 21, 2026, the alert assigns the flaw a CVSS 3.1 score of 9.1 and covers Oracle Java SE and GraalVM.
The update blocks affected intermediate certificate authority certificates linked to eMudhra. According to Oracle’s security advisory, the issue affects the Java Secure Socket Extension, or JSSE, component.
Oracle rates exploitation as low complexity, with no authentication or user interaction required. Successful attacks can expose or change sensitive data, although the risk matrix lists no availability impact.
#cybersecuritynews #vulnerability
9
530
you only need a coding agent and the right API's, don't let the tech geeks fancy word scare you
the average affiliate deals with more complexity daily than the average developer
Here is how I am using GrokBot as a media buyer managing over $200K/day in Affiliate and Ecom Ad Spend:
Truthfully, I didn't expect my previous post to get so much attention, so I decided to actually share some value (hopefully).
Basically I've got all APIs plugged into a handful of pots: Google Ads, YouTube, Everflow, Lead Prosper, and Google Sheets
Affilate:
- Checks my margin every hour. It matches my ad spend to the revenue I actually got paid, not the conversions Google reports. I match back campaignID and creativeID for real time tracking
- Tells me what to cut and what to scale. It follows my own rules, like cutting after 3 losing days and only raising budgets at certain hours. I've also had it run a study on my media buying actions over the last 60 days to see how my budget increases effect margin, so now I have predictor of how aggressively I should raise and at what hours of the day are the best to make changes
- Flags losing ads early. If an ad spends too much with no leads, I know before it burns real cash. For example if it spends 3x target CPA it automatically flags to be paused
- Judges ads on lead quality, not just lead count. It matches every ad to which leads actually turned into sales, signs or appointments. I can pass back appointment rate, sales rate, call length, revenue per call, etc to each ad and transcript
- Predicts where the day will end up, and how sure it is, so I'm not guessing at noon. Since there is a time delay between click and when the revenue fires back, sometimes I can be spending several thousand dollars per hour and see major drops in CVR, the predictions help me decide if the traffic quality is good or bad before the revenue fires back
- Tests my landing pages and videos against each other and shows me which ones actually make money. I can see my profit per click per landing page, sure Redtrack does this, but I use Everflow not RT which doesn't have a built in LP tracker
- Checks what I changed before it gives advice, so it doesn't tell me to undo something I just did. I make sure its constantly learning from MY behaviors not making decisions on its own
- Updates my P&L sheet every day, so I never build reports by hand. I created a Google sheet and it pulls spend from Google Ads and Revenue from Everflow and creates a P&L daily of each offer and each vertical
- Pulls YouTube transcripts and ties back performance to the script and hook used in each ad
- Shows all my bank and card balances in one place.
Matches what each partner owes me against what actually hit my bank, so I see who's paying late. This is super helpful because it instantly tells me where my money is and knows which advertisers pay on which payment terms, which credit cards are due when, and helps me manage and understand my cash flow
E-com:
- Ecom is a lot more robust and I am still building out the flows but I will mostly be using it for understanding my media buyers behaviors, for example, if there are campaigns running hot in client accounts, I get a notification to my phone alerting me they haven't made a change they should have made
- The same goes for scaling, one of the hardest things is to teach buyers how to SCALE and I mean really SCALE not just 20% raises here and there. These alerts ping to me when we identify a campaign that can scale and if they don't make the changes or make changes against how I would do it, I get a notification which provides a teaching moment to the buyers
- Of course, we are doing high level script and hook analysis on our creatives and landing pages
- It also sending notifications when certain campaign types are underperforming or if it identifies new opportunities
Overall, there is A LOT I am still learning and I've only had this bot set up for about a week, but its truly a revolutionary approach to buying and management. Sure, I should have been using Hermes agent a long time ago, but I am not the most tech savy person and have been quite busy, but GrokBot was SO easy to set up, I truly believe anyone and everyone can use it and improve massive amounts of workflows with it.
Any questions? LET ME KNOW in the comments
1
10
797
Whop is leagues above other companies in becoming agent native.
A month ago customers suddenly lost access to a web app. Some authorization bugs on the server. I gave a coding agent the whop cli and it migrated everyone on the whop frontend and matched the fixes on the backend.
Today, Whop is introducing Economic Intelligence, a decision making system built from business actions on platform.
Billions of people should have the ability to build profitable businesses.
Economic Intelligence answers "what next?" for businesses at all levels.
13
903
With herdr.dev you can still do this with any agent. spawns and orchestrates it using the herdr api
I still prefer to use sonnet for copywriting. Fable for audits or to solve a problem for Sol or FlashV4
5
782
a coding agent can do this for you in less than 3 minutes. if you provide it with API's to scrape whatever you're scraping.
just use helix-db.com takes care of everything else.
get one good web search API and your agent can figure out what API's you need and which are the best ones for your usecase.
I do this whenever I have a niche specific problem I want to quickly solve
For example I used this for when I wanted to figure out the best and cheapest way to do video gen at scale. I sent one prompt then 40 min-ish later I was generating seedance lvl videos at 20x cheaper.
i extracted all the lessons from my 14,000 tweets & turned them into a browsable knowledge-base...
you can read them here: eptwts.com
you can use claude to write copy, just not in the web app. its less cucked if you use claude code. Also as with everything else, the model matters less than everything around the model you pick
-references/swipes
-frameworks
-data sources
-context management
a shittier model using a better harness will smoke claude anyday. if they have the same harness the difference between them will still not be as big as you would think.
How is everyone writing copy now a days?
Claude seems to be super restricted on writing anything remotely aggressive or where claims just cross over the slightly exaggerated line
11
1,122
your daily life and work will improve drastically by just giving your agents apis that it can use to get data. Scraper apis (rapidapi) search apis (tinyfish, exa) and then apis for specific work u do. a lot of people already use the search apis, but having the ability to scrape and get the exact data for whatever you need is underrated.
I generate 50k images for ads a month and it costs me less than i pay for twitter premium. I did not come up with the method behind that, my agent did with 1 scraping api and 20 min of research.
the number 1 thing in common with anyone who is cracked at leveraging agents and actually gets value from it, is being able to get data and the usage of said data.
whatever problem you have or task you are trying to get done or bottleneck you are facing, you can solve with just getting the right data. Get scraping apis and get whatever data source that's needed for your daily work.
3
14
690
if you are using coding agents like codex or claude code and your setup does not include deepseek v4 flash, your opinion on anything is irrelevant as you are a retard
tell your coding agent to setup deepseek to be available by searching up the most popular ways of using it by the community
takes you zero setup time and you unlock your ability to just throw agents at every problem and get shit done way faster
7
812
you can run the new Qwen model that beats Opus on benchmarks with basically unlimited FREE usage right now.
you only need to:
- sign up to modal, u get 30 free credits every month (more than 24 hours of unlimited access)
- ask your agent to deploy using the modal cli
- wait 5 min
6
574
Opus4.6 perfomance claims have to be verified immediately, deploying and gonna update with news
unless it goes really well. deleting if thats the case
We promised open weights for Qwen3.8. Now, time to meet them! 🎉
⚡ Qwen3.8-27B:
- A native multimodal dense model. With just 27B parameters, it outperforms Qwen3.7-Plus overall and shines in real-world coding & office workflows.
- 262K native context, easily extendable to 1M tokens via YaRN.
- Built for builders. Highly efficient, high-quality, and licensed under Apache 2.0.
🚀 The open weights for Qwen3.8-2.4T-A95B (Max-level) have also been released recently.
Whether you're shipping lightweight applications with Qwen3.8-27B locally or building agents with Qwen3.8-2.4T-A95B, they're yours now!
Download, deploy, and build something we haven't imagined yet. 👀👇
- Hugging Face:
huggingface.co/collections/Q…
- ModelScope:
modelscope.cn/collections/Qw…
5
537
the ceo of shopify is properly lvrging and the ecom niggas are paying 30x markup on AI UGC because they are intimidated by the terminal
my favorite way to explain people herdr is to just open a new workspace, launch an agent and tell it to:
read `herdr --skill` then split 10 other panels that make my computer look like some scifi hacker terminal from the movies
4
645




















