Futureformed (@Futureformed) header
Futureformed (@Futureformed) profile picture

The future will be formed by an alliance of human and technology. Choose the right partner on the AI journey. AI expertise, responsibly done.

Joined August 2026
Photo by Futureformed (@Futureformed)Photo by Futureformed (@Futureformed)Photo by Futureformed (@Futureformed)Photo by Futureformed (@Futureformed)Photo by Futureformed (@Futureformed)
Futureformed retweeted
can’t believe they built the pyramids without MCPs
56
89
9
1,009
29,290
Futureformed retweeted
What's the hell is going on at Anthropic
138
97
28
4,174
571,188
Futureformed retweeted
give a man a fish and you feed him for a day teach a man to Claude Code and you financially ruin him with distracting side projects for the rest of his life
419
1,672
202
25,784
638,525
Might want to take a look at this @firecrawl
37
Great suggestions from one of the smartest guys I know
Good prompt below to do some basic hardening and keep the zombie swarms off the walls for a little bit longer. Not definitive, and we're fucked anyways... but. ‘You are my security engineer. Help me harden my machine, my code and my home setup against the attacks that are common now: compromised packages, leaked keys, forgotten servers, and agents with more access than they need. Don’t treat this as a checklist. Work out what I actually run, then think about where someone would get in. Everything below is a starting point, not a complete list. Map me What’s on this machine and in my repos: projects, languages, package managers, where each thing deploys, which keys and accounts it uses, which AI agents and MCP servers I run and what they can reach. Ask me about anything you can’t see, like my router or which accounts matter most. Find my weak spots Where would an attacker get in, and what would they get? Rank by likelihood and damage. Hunt the edge cases in my own setup: an app that deploys straight from main, a key an agent can read, a token that never expires, a secret in a public repo’s history, a server I forgot I’m paying for, a tool I installed once and never updated, anything listening on a port. Places to start Dependencies, the main way in. Known bad versions, signs one already ran, lockfiles, install scripts, how fast new releases reach me. Secrets. Plain-text keys in dotfiles, configs, shell history, scripts and git history, and .env files committed to public repos. Wallet keys and contract-owner keys first. Spend caps on every paid API. Agents and tools. MCP servers, plugins and global packages that pull the latest version on every run. GitHub and hosting. Dependabot, dead repos, deploy keys, Actions secrets, what’s running and what it costs, long-lived tokens. Agent permissions. What my agents are allowed to run and read without asking. Laptop. Firewall, what’s listening on a port, folders on my PATH that anyone can write to, old tools I forgot about. Key accounts, home network, backups. Make it hold Fix what’s safe, then set up the thing that keeps it fixed. Lockfiles and a delay on new releases. Grouped Dependabot security PRs. Spend caps. Tokens scoped to one job, with an expiry date. Keys in a keychain or password manager, not dotfiles. Secret scanning on push. A way I’d find out if something went wrong. A backup you’ve actually restored from. Rules Read first. Change nothing until you’ve shown me what you found. Never print a secret. Don’t cat or list config and key files. Use grep to find key names and print only the name and file:line. Ask before anything destructive, anything that costs money or anything that touches production. Don’t install new tools without asking. If a scanner like gitleaks would catch what you can’t, ask me to install it. Be thorough. Check every repo and every hosting account, not a sample. If you run out of room, say what you skipped. Check every fix worked. If you can’t check something, say so. Save it. Keep a living SECURITY.md in a folder I choose, or in this project’s memory. Done, do now, later, and what changed since the last run. Never write a secret into it. Offer to schedule a re-run every two weeks. Each time, read it first, recheck what’s done, look for anything new, then carry on.'
17
Replying to @Futureformed
So cool, privacy, and no need to remember to track.
1
1
12
Got a walking route but don't know how many steps it is and how many calories you'd you burn? Introducing gStep - a browser extension that's private & FREE. It looks at the route and tells you how many steps and calories are involved. Download now! gstep.fit/
1
2
33
Last month at @RamenClubHQ I’ve been building a proper member portal for our community to navigate all our resources, all built with Codex desktop app: - stack is Cloudflare Workers, Supabase db/auth, Loops for email, Stripe for payments, Astro frontend - added full billing automation for existing and new members (probably took the longest out of everything) - all 500 discounts, plus editable member directory, recorded content, upcoming events here via Secret, Drive & Luma APIs - PostHog & Fathom analytics & session replays - automated verification between account and Slack profiles upon joining - all fast, secure, works on mobile Next: - member directory 2.0 - make all browsable via MCP - members can submit own discounts, apply to host live classes, apply to host events, volunteer in other ways - education 2.0 - some other useful and fun stuff 😉 (@HCJMartin has been cooking too) We’re just getting started.
15
1
2
34
2,098
Another really smart product and founder. Lots of agent tools out there but few truly useful one - @flocker_md is one of those.
Grok, Claude and Codex work EVEN BETTER when they work together✌️ Super excited to be launching Agent Profiles on Product Hunt Monday! We've tuning flocker.md for 10 months! Unlock 2 months unlimited live AI profiles and task orchestration FREE👇 link in comments
3
22
Love the simplicity of this. So many tools / apps / platforms to choose from, but they all miss the key ingredient - keeping it simple. Salute @_baretto !
For 5 years, Tiiny did one thing: You bring the files. We put them online. Today that changes. Introducing Tiiny CREATE. No empty prompt box. No guessing what to type. We guide you step by step → you get something great → it's live in seconds. 🚀
1
3
76
Seen this at work and it’s super impressive. Don’t sleep!
Looking for a dashboard to manage all your AI agents? That you can access anywhere!? What are you waiting for? 👇 flocker.md/#ask-flocker Flocker gives all your AI agents a unique (private) profile page — Claude, Codex, Opencode, Hermes... @ClaudeDevs #AITools #Automation #NoCode #BuildInPublic
3
45
Futureformed retweeted
Ever wanted to give your AI agents a live home page? Wished you could magically transfer your agent's behaviour to another platform? Flocker Agent Profiles is launching on Product Hunt on Monday 21st of September! Track us 👉 flocker.md/#mail-sub
2
1
228
Why have the Claude UX team not managed to find a way to mass clean up sessions in CC macOS? Rn you have to go into each one - per project - and archive them? If you open new sessions to manage context, it becomes a mess. @trq212 you able to comment pls?
2
20
In case you missed the past few weeks of #Busmaxxing It culminates in a delightful evening at Ramen Space tomorrow for Shipaton w/ @RevenueCat RSVP here: luma.com/ramen-cl5p We'll have some great talks, app roasts, and announcing our grand winner. See you there! #StrictlyBusness
PARKED UP OUTSIDE RAMEN SPACE STRICTLY BUS-INESS
1
2
23
1,451
The AI CEOs are demonstrating one of the core WINNERS TAKE ALL moves: By naming, and framing their way, a social problem they’ve also caused, they appoint themselves leaders of the search for solutions when they should be asking for the society’s mercy.
29
516
25
1,742
44,507
new album just dropped
Replying to @charlierward
Yet another earlier. Being stalked by buses. Buses Rule Everything Around Me (B.R.E.A.M)
2
1
5
570