Adam Paciorek (@AdamPaciorek) header
Adam Paciorek (@AdamPaciorek) profile picture

Award Winning #DigitalMarketing Consultant working with innovative #StartUps.

London
Joined July 2009
Photo by Adam Paciorek (@AdamPaciorek)Photo by Adam Paciorek (@AdamPaciorek)Photo by Adam Paciorek (@AdamPaciorek)Photo by Adam Paciorek (@AdamPaciorek)Photo by Adam Paciorek (@AdamPaciorek)Photo by Adam Paciorek (@AdamPaciorek)Photo by Adam Paciorek (@AdamPaciorek)Photo by Adam Paciorek (@AdamPaciorek)Photo by Adam Paciorek (@AdamPaciorek)Photo by Adam Paciorek (@AdamPaciorek)
This is insane. Hermes is making films on your own machine now. 🎥 A trailer. A product ad. A music video. A short film. You pitch it in chat. Hermes shoots the clips, casts the voices, scores it and adds the sound effects. Directed by Hermes. Imagination Studio for Hermes plugs Hermes into a full AI film studio running on your own hardware. From any chat, Hermes makes: ▸ Pictures ▸ Video clips with sound ▸ A voice line in any voice you describe ▸ Songs ▸ Sound effects Every file lands back in the conversation, and long renders run in the background while you keep going. Works from Hermes Desktop, the terminal or Telegram. Inside the Studio itself you can write the script, plan every shot, pick the look and cut the film. It all renders locally, from DGX Spark and RTX PCs to Apple Silicon Macs, and your files stay on your machine. Setup: install Imagination Studio on your machine, add the plugin from the official Hermes catalog, then ask Hermes. Built by @u1tra_instinct and @SamJWasserman.
32
40
10
442
33,583
Adam Paciorek retweeted
Ein praktisches neues Plugin für die Auswertung von Social Media Multimedia Inhalte
THIS IS SO SICK One long video in. A week of content out. Content Engine is a new community plugin in the Hermes catalog. Paste a YouTube link and Hermes finds the best moments for you. What it does: ▸ Pulls the video's timestamped transcript ▸ Scores every moment on hook, completeness, standalone value and visual payoff ▸ Picks the top 45 to 60 second clips, with no overlaps ▸ Writes first-draft copy for X, Facebook, YouTube Shorts, TikTok and Instagram ▸ Cuts the clips with subtitles burned in, horizontal and vertical No API keys. No account. No editing software. Got a podcast, a stream or a long tutorial sitting on YouTube? That's a week of posts waiting to happen.
2
1
739
Got a "can Hermes do this?" question? JUST ASK HERMES. It ships with a built-in hermes-agent skill that knows where the answers live. For any "can Hermes do X?" or "how do I set up X?" question, it starts from the official docs index and goes to the page that answers. It's even told never to say "Hermes can't do that" from memory. It checks first. Try it: ▸ /hermes-agent how do I run a task every morning? Cite the docs page. ▸ /hermes-agent can you control my browser? Link the page. ▸ /hermes-agent which memory providers can I plug in? Want ideas instead of answers? The docs also have a User Stories & Use Cases page: real community setups sorted by category, from dev workflows to personal assistants and business ops. Fewer guesses. Every answer comes with a page you can check.
5
8
41
4,057
Adam Paciorek retweeted
Polish🇵🇱 heroes: the surgeons who rushed to help the wounded police officers Two outstanding Kraków surgeons, Prof. Michał Pędziwiatr and Dr. Mateusz Wierdak, were having lunch at a restaurant near the site of yesterday’s tragedy on the Main Square. When they learned that two police officers had been seriously shot while detaining an attacker, they immediately abandoned their meal and rushed to help. One of the officers was in immediate danger of death. Professor Pędziwiatr not only helped the wounded at the scene, but later also operated on one of the officers at the hospital. Their lightning-fast reaction may have been crucial to saving the officers’ lives. Pass it on, let it spread 🇵💪
104
862
37
4,540
76,427
Okay, this one's next level. 🎨 HERMES CAN NOW MAKE IMAGES WITH NANO BANANA AND VIDEOS WITH GEMINI. 🤯🎬 Gemini Media is a community plugin in the official Hermes catalog. It plugs Google's Nano Banana and Gemini Omni into Hermes's own image and video tools. What it does: ▸ Text to image, plus edits guided by up to 14 reference images ▸ Images in 1K, 2K or 4K ▸ Text, image or reference to video, with native audio ▸ 3 to 10 second clips, landscape or vertical ▸ One Gemini API key covers both What you can do with it: ▸ Turn a product photo into a clean ad shot ▸ Animate a still image into a short clip ▸ Make a vertical clip for a post from one prompt Ask in chat. Get the image or the clip back.
10
27
3
377
19,443
Your AI agents need a project manager. Hermes already has one. 🔥 Hermes Kanban turns your projects into boards your agents work through on their own, and you can see where everything is at a glance. ▸ A board for every project, as many as you want ▸ Hermes splits a rough idea into tasks for you ▸ Each task goes to the right agent ▸ Agents can see what's blocking what, and pick up the moment they're clear ▸ Blocked shows you exactly what needs your call ▸ The finished work comes back on the card It's built into Hermes Desktop. Most people just haven't switched it on. Watch one idea turn into four cards and a finished guide 👇
16
15
1
203
12,376
Adam Paciorek retweeted
Dwóch policjantów, którzy obezwładnili napastnika podczas zamachu w Krakowie, zostało ciężko rannych. @K_Stanowski uruchomił zbiórkę na ich wsparcie. Pomóżmy! ❤️ 👉 zrzutka.pl/bohaterowie
13
73
5
470
64,894
Adam Paciorek retweeted
Tired of explaining your setup to 𝗛𝗲𝗿𝗺𝗲𝘀 again every new chat? 🧠 It can remember you. 7 tricks, one room each in the video: 1. Memory budget: MEMORY.md holds 2,200 chars. Run /new when a task is done. 2. A diary on you: say "Remember I like short answers." It lands in USER.md. 3. It cleans itself: a note over the limit is refused until old ones go. 4. Old chats: "Search my old sessions for how we fixed the login." 5. A room per project: hermes profile create work --clone 6. Your Obsidian vault: the built-in Obsidian skill reads and writes your notes. 7. A backup of its brain: hermes backup --quick Start with 1, 4 and 7. Save this, and every step is in the article below 👇 Which one are you trying first?
6
5
1
59
4,550
Hermes bots have coworkers now. 🤖🔥 HERMES BOT GROUP CHATS NOW RUN THEMSELVES. Drop a message that names nobody and the room picks who takes it. "Assigned to Builder" shows up before anyone replies. Out now in Hermes v0.23.0. What it does: ▸ The room picks who answers, or you give it a lead bot that hands out the work ▸ Bots pull each other in, and you see every handoff ▸ Each bot shows what it's doing live, like "scout · terminal · 7s" ▸ Scheduled jobs post into the room and wake only the bots they @mention ▸ Every room gets a purpose and shared notes each bot reads on its turn ▸ Bots from different machines can share one room ▸ Long discussions pause at the limit. Hit Continue to keep going What you can do with it: ▸ Post a bug and let the room hand it to your builder bot ▸ Have a morning job drop the news in the room and wake only your writer bot ▸ Put your project brief in the shared notes so every bot works from the same plan ▸ Let a scout, a builder and a critic talk a plan through, then hit Continue if they need more rounds Your bots stop being a row of separate chats. They work like a team now.
16
46
7
637
52,020
Every other agent harness is cooked. 💀 HERMES AGENT v0.23.0 PUTS YOUR WHOLE DESKTOP ON YOUR PHONE. 📱 The Hearth Release is out. Your full Desktop workspace runs in any browser, and one gateway now owns every chat. Same conversation on every screen. What's new: ▸ Desktop in any browser, phone included. Turn on Web Access in Settings ▸ One gateway: CLI, TUI, Desktop, bots and cron all share the same chats. No more doubled or missing messages ▸ Every machine in one Desktop window. Jump between this computer and your SSH boxes instantly ▸ Esc twice (or /rewind) to edit and resend an earlier message, or fork a new session from it ▸ Pasted images and screenshots draw right in the terminal ▸ Paid tools show their price per call, and Hermes asks before a chat goes over its limit. Even in YOLO mode ▸ Group Chats can pick who answers on their own, and cron jobs can post into a room ▸ Plugins update in one go, and anything that would break your setup gets held back ▸ The plugin catalog jumped from ~600 to 1,050+ ▸ Lighter all-day runs: one-shot calls send ~1,570 fewer prompt tokens, and cron jobs share a prompt cache ~1,950 PRs from 666 contributors. Name another harness shipping like this. hermes update and it's yours.
30
57
7
717
45,267
Adam Paciorek retweeted
Hermes merged 47 PRs on November 4 The ones worth knowing: - Bot Screen lands: each bot gets its own Xfce desktop streamed into Hermes Desktop, and you can take over and hand it back - One unified gateway now owns every local session: classic CLI, TUI, Desktop, API, ACP, bots and cron - The gateway can now serve the Desktop workspace straight to the web app - Cron gets serious about time: per-job timezones drive the due-scan, no more firing an hour early on DST fall-back, and every scheduled prompt knows when it ran - A full or read-only cron store no longer stops the scheduler, jobs catch up once on recovery and the outage shows up in status, doctor and metrics - hermes update now backs up HERMES_HOME automatically before it runs - A killed Windows updater no longer strands paused gateways or resumes them onto a half-written tree - Claude on Bedrock gets the full 1M context, and missing Bedrock models are sized from their AWS cards - Email now requires an authserv_id pin, so forged dmarc=pass headers are refused - Hermes Talk, progress-squash and telegram-stickers join the plugin catalog - Desktop ships Electron 44.9.3 without breaking fresh Windows installs, plus faster Windows startup Here is the full list: DESKTOP IMPROVEMENTS (10 PRs) - feat: Bot Screen, per-bot Xfce desktop streamed into Hermes Desktop, take over and hand back - feat(webapp): serve the Desktop workspace from the gateway - feat(desktop): native in-app downgrade: chargeless preview, schedule, undo - fix(desktop): upgrade to Electron 44.9.3 without breaking fresh Windows installs - fix(desktop): Windows SSH probe parses again, try/catch no longer split by ';' - perf(desktop): reduce Windows startup I/O and blocking checks - fix(desktop): a hidden window no longer drops its own healthy gateway socket - fix(webapp): a Web Access Start stays visible until the gateway serves it - fix(linux): hermes.desktop Exec no longer flips to the module form on a cold-context launch - fix(cli): Linux hermes.desktop entry launches instead of silently dying on system python DEVELOPER TOOLS & PLUGINS (16 PRs) - feat(plugins): add Hermes Talk to the plugin catalog - feat(plugin-catalog): add progress-squash - feat(plugin-catalog): add telegram-stickers - fix(plugin-catalog): re-pin digital-marketing-pro to 3.57.8 (model router, registry-refresh keys in a header) - Add simple terminal - fix(tools): make managed Node authoritative in the MCP stdio PATH - feat(local-models): calibrate Spark recommendations and default 27B reasoning to medium - fix(bedrock): 1M context for current-gen Claude on Bedrock - fix(bedrock): size missing models (GPT-6.1 Terra) from their AWS cards and label fallback context windows - feat: component-separated logging with session context and filtering - feat(nix): container-aware CLI, auto-route into managed container - fix(nix): switch nixpkgs input from nixos-24.11 to nixos-unstable - feat(update): auto-backup HERMES_HOME before hermes update - fix(update): gate the fork-upstream prompt for --yes and non-tty runs - fix: surface self-improvement review summaries across CLI, TUI, and gateway - Code health can only improve: per-unit CC/size ratchet, Hermes invariant rules, scripts/check SECURITY & SECRETS (6 PRs) - Email requires an authserv_id pin; forged dmarc=pass refused (behaviour change) - fix(deps): clear release-age-eligible security alerts (undici, zod, fast-uri, qs, pypdf, pyasn1, tornado, filelock, setuptools) - fix(approval): catch hermes gateway stop/restart behind a profile flag - fix(cron): don't strict-scan script-injected output in no-skills jobs - chore(deps): bump idna from 3.11 to 3.15 - chore: update base debian images to 2026-10-20 INTEGRATIONS & CHAT (2 PRs) - Unified gateway runtime: one gateway owns every local session (classic CLI, TUI, Desktop, API, ACP, bots, cron) - fix(matrix): bind PgCryptoStore device_id so fresh E2EE installs work STABILITY & FIXES (13 PRs) - fix(cron): a full or read-only cron store no longer stops the scheduler; jobs catch up once on recovery and the outage shows in cron status, doctor, metrics and the home channel - fix(cron): per-job timezone must drive the due-scan, not just compute_next_run - Cron jobs with a timezone no longer fire up to an hour early during the DST fall-back hour - Cron jobs know what time they fired: run time in every scheduled prompt - fix(cron): execution, delivery and incident history ordered by instant, not by text - fix(agent): include timezone and UTC offset in system prompt timestamp - fix(update): a killed Windows updater no longer strands paused gateways, and never resumes them onto a torn tree - refactor(process): one owner-death supervisor for every long-lived Hermes child, plus a ratchet against new detached spawns - fix(onboarding): import Path in methods_onboarding (serve startup NameError) - fix(ci): repair red main: busy-mode test + missing checkout in skills-index workflows - test(cron): profile-timezone test no longer fails the day before a DST change - test(telemetry): settle the export worker before engagement assertions read the store - test(delegate): fix race in test_finite_batch_returns_parent_interruption 47 new features and fixes teapawt.org/iamlukethedev/status/2…
Hermes merged 54 PRs on November 3 The ones worth knowing: - Cron jobs can now post into Group Chats and only wake the bots they @mention, and rooms get a purpose and shared notes - Group Chat rooms now look like the native chat and show each bot's live state - Opt-in lead routing lands for Group Chats - Bot group chats can seat bots from different machines again - Pin Kanban boards as quick-switch tabs in Desktop - A failed Desktop update now shows what went wrong, with a support link in the same box - The Claude subscription plugin lists your account's own models, Mythos 5.1 included, and Claude Mythos is treated as mandatory thinking - Claude subscription errors now include the upstream request ID - Otter.ai MCP OAuth login works even when its issuer differs only by a trailing slash - The Windows installer no longer gets stuck on a half-installed Git - Approval rules got tighter around credential uploads and stray variation selectors - A batch of security alerts cleared across undici, zod, qs, pypdf, tornado and more Here is the full list: DESKTOP IMPROVEMENTS (13 PRs) - fix(desktop): keep one tool row per call when events repeat or ids are reused - fix(desktop): keep notarizing through dropped connections to Apple - feat(desktop): Web Access Start shows its step and elapsed time - Desktop no longer warns that a closed local app's MCP server failed its health check - Desktop: a failed update shows what went wrong and gets a support link in the same box - fix(desktop): negotiate the session-authority wire on public remote connections - fix(desktop): move macOS traffic lights closer to the window edge - feat(desktop): pin Kanban boards as quick-switch tabs - fix(desktop): support older gateways in Quick Entry - Windows installer no longer gets stuck on a half-installed Git ("INSTALL DIDN'T FINISH") - fix(bundles): refuse a macOS update ZIP with entries lacking owner write - fix(bundles): bake pycs owner-writable so Squirrel.Mac updates install - fix(bundles): keep uv's wheel links from the payload cache to the machine cache VOICE & REAL-TIME (3 PRs) - fix(discord): probe for libopus once, off the event loop - fix(setup): offer Edge TTS voice selection without resetting the current voice - test(install-e2e): journeys turn STT off so an old release's boot probe cannot stall the send DEVELOPER TOOLS & PLUGINS (14 PRs) - Optional skills that the MCP catalog, catalog plugins, built-in web search or vendor skill repos cover are removed - Catalog installs of the Claude-subscription plugin now list the account's own models, Mythos 5.1 included - fix(pm): bind the core package for plugin extras that name it - fix(pm): a virtual plugin's own-extra references follow its workspace rename - fix(pm): leave out a carried extra that no longer builds instead of failing every update - fix(pm): default a CMake policy floor into uv source builds so python-olm builds under CMake 4 - fix(pm): plugin dev/test extras and dependency groups no longer fail the update lock - fix(pm): build native sdists with gcc/g++ when the pinned Python's clang is missing - fix(pm): `pm install --without EXTRA` drops a carried extra (matrix) and keeps it dropped - fix(pm): a tooling-only plugin pyproject no longer hides plugin.yaml dependencies - fix(pm): ignore desktop metadata files without hiding package payloads - test(pm): patch registry package classes, not their singletons - fix(hooks): expose the owning Kanban task id in hook payloads - chore(plugins): pin birkin-mnemosyne to core 0.9.0 SECURITY & SECRETS (4 PRs) - fix(deps): clear release-age-eligible security alerts (undici, zod, fast-uri, qs, pypdf, pyasn1, tornado, filelock, setuptools) - fix(approval): an approved rule no longer hides a credential upload - fix(approval): flag variation selectors that do not follow an emoji - fix(approval): credential-upload rule only counts curl/wget upload flags INTEGRATIONS & CHAT (10 PRs) - feat(bots): cron jobs post into Group Chats (wake only who they @mention); rooms get a purpose and shared notes - feat(bots): group rooms look like the native chat and show each bot's live state - feat(bots): opt-in lead routing for Group Chats - fix(desktop): Bot group chats can seat bots from different machines again - fix(bot-mode): relaunch no longer pins "Hermes gateway is not connected" in the Bots pane - Hermes Cloud: the Telegram card points to setup that works on Team Gateway (Portal linking for managed agents, /link codes otherwise) - fix(mcp): Otter.ai OAuth login completes when its issuer differs only by the root trailing slash - Claude subscription errors now include the upstream request ID (catalog re-pin) - fix(anthropic): treat Claude Mythos as mandatory-thinking - fix(gateway): deliver settled replies before releasing turn media STABILITY & FIXES (10 PRs) - fix(session): browse recent sessions with a ten-result default - fix(delegation): reserve child transcripts before publishing session IDs - fix(profiles): keep existing profile dirs when a force install fails - fix(web): run config schema build and egress status read off the event loop - test: serve external services from fixtures and fence every off-machine host - ci(js): drop vitest's run history from the cached node_modules - ci: remove the Intel macOS CI and release legs - ci: run macOS arm64 bundle, smoke and install E2E jobs on Blacksmith - perf(ci): fork Linux test files from a preloaded pytest server, 3 Linux slices, warm Windows PowerShell and Git caches - docs(ci): document every workflow's triggers, skips and release paths 54 new features and fixes teapawt.org/iamlukethedev/status/2…
4
1
1
23
10,842
😍
You can now hire an AI team that clocks in on its own. Strawberry just launched app triggers, powered by Jev from @typesafeai. Your companions start the moment something happens in your apps. No re-prompting, no babysitting. The team: ▸ Sales Sally finds qualified leads and writes the outreach in your voice ▸ Talent Theo sources candidates and scores each one against your needs ▸ Executive Ellis handles your morning brief, drafted replies and follow-up reminders ▸ Marketing Martin writes posts for each channel based on what's worked for you ▸ Consulting Cora turns client calls into proposals What sets them off: ▸ New email from a lead → Linear task, HubSpot contact, notes in Notion ▸ Your weekly sync ends → notes and follow-ups are ready ▸ New Slack message in a channel → follow-up tasks You decide what each one can do without asking. The plans are literally called Intern, Part-time and Full-time. Start free at strawberrybrowser.com 👇 @strawberry
1
2
920
Adam Paciorek retweeted
Try this prompt in @claudeai with Apify MCP: “Pull [competitor]’s Meta ads, keep the ones running 90+ days, and list the landing pages they send traffic to.” That’s step one of @TheMattBerman’s method for finding the page change behind a competitor’s best ads 👇
your biggest rival already A/B tested their landing page. with their money. you can see the results for free but most people only look at half of it. the ad library shows what's running. it DOES NOT show why it works. the why is on the page the ad sends to. and the wayback machine has every version of it. here's what to do: 1. pull the ads scrape their meta ads with any @apify ad library scraper. you want start date, copy, and the landing URL for every ad. 2. keep the long runners filter to 90+ days. 60 is too soft. a lot of ads still die after that. 90 day ads are the ones paying. 3. find the money angle have claude group them by hook structure (not topic). the group with the most 90 day ads is the angle making them money. 4. map ads to pages pull the landing URL for every long runner. one page usually serves a bunch of ads. 5. pull the history run each URL through the wayback machine api. store the timestamp + the above-the-fold content (H1, offer, price, guarantee, proof). one snapshot a month, last 12 months. 6. find the change use claude or jev to compare month to month snapshots. look for the month the top of the page changed in a real way. then check the ad start dates. if the long runners started right after, that's your lead. if the page changed and the ads died, that's not the move. flag the month where the above the fold changed in a big way like a new headline angle, stronger offer, price shift, new guarantee, more visible social proof 7. test the move build one page that makes the same structural change (offer, guarantee, headline angle) but you still use your own language/product positioning. run it against a control as a new ad set (one URL per ad set so you dont reset learning). track your tests with something like this: money angle: page change before the long runners: month: ads that started after: my test: if we ___ on our page then CPA should drop try it on one brand. you walk away with a test + takes just an afternoon.
3
4
15
1,323
Adam Paciorek retweeted
Don't let your agents become hostages to a single company's policies For the Grok Bot users who are unhappy with random (and unannounced) usage limits, there's now an easy way to switch to Hermes: hermes migrate grok-bot Bring all of your memories, skills, and routines to Hermes with just a couple commands Now if one AI lab changes their usage policies, you can just easily switch providers I made a video showing this process and how easy it is. Check it out!
Replying to @DogecoinNorway
The cost of serving Grok Bot was unsustainably high relative to revenue
13
21
2
183
12,003
😍
Your agent can set up its own computers now. Ask for 5 and it puts another agent to work on each one. Here's what that could look like. You run a small web design side hustle and need leads. You ask your agent to find local businesses with outdated websites in 5 cities. It spins up 5 @Orgo computers from the Hermes template, one per city. Each agent opens a real browser, checks the sites, notes what's broken and saves its list. You can watch any screen live, or take over if one gets stuck. When the work's done, your agent collects the files and shuts the computers down. Why give every agent its own computer? ▸ Its own browser, terminal and files, on Linux or Windows ▸ Ready in seconds from a template ▸ All of them work at the same time ▸ They keep going with your laptop closed ▸ No hardware to buy It works from Claude Code, Codex, Cursor or any MCP client. Every agent needs a computer. Now your agent can set them up. Easiest way to try it: connect Orgo to your agent and ask for 3 computers. Use the link below for a free 3-day Orgo trial: orgo.ai/r/hermeswatcher
1
2
796
The most slept-on feature in Hermes Agent is sitting right next to your chat. It's a real browser built into the desktop app, and Hermes drives it while you watch. What it does: ▸ Opens any site, your localhost dev server or a local file ▸ Clicks, types and scrolls on its own while you watch its cursor ▸ Marks what it finds with labels right on the page ▸ Comment mode: click any element or drag a box, then leave a note ▸ Every comment reaches Hermes with a screenshot and the element's selector ▸ Console and DevTools right in the browser bar What you can do with it: ▸ Have it find a video and play it in 4K, right beside your chat ▸ Ask it to pull the top results off a page and label them for you ▸ Fix up a site you're building: pin what bugs you, Hermes makes the changes Watch it work. Point at anything.
17
21
3
277
34,538
Adam Paciorek retweeted
Hermes can now import your whole Grok Bot setup with one command @Teknium just merged it into hermes-agent main: - hermes migrate grok-bot - your routines become Hermes cron jobs, same schedule and time zone - skills, memories and your bot's role come with them - routines arrive paused, so nothing runs twice while you switch - no logins handed over: each bot writes its own export It also works for OpenClaw, Claude Code, Codex, Cursor and OpenBot. Just run hermes migrate and it finds what you have.
34
12
8
134
8,294
Grok Bot users got rug pulled on usage limits. Hermes just built the exit. hermes migrate grok-bot Your whole Grok Bot setup moves into Hermes: ▸ Scheduled routines become Hermes cron jobs ▸ Skills come over as Hermes skills ▸ Memories merge into Hermes memory ▸ Your bot's role and rules carry over ▸ Apps with an MCP URL get wired in as MCP servers How it works: ▸ Hermes gives you an export prompt ▸ Paste it into Grok Bot and it writes its own export file ▸ Run the command and Hermes picks it up Routines arrive paused so nothing runs twice. Turn them on when you're ready. Fresh installs ask during setup if you're coming from Grok Bot. Works on Mac, Windows and Linux, and never touches your logins. Lands in Hermes today. Same routines. Any model you want. Your own machine.
88
87
46
1,089
86,575