Tramch (@Tramch) header
Tramch (@Tramch) profile picture

🇱🇧-🇩🇪 • #Gaming • #Digital #Marketing - #Advertising - #Comms • #Apple • Views are strictly my own • Repost/Like/Share ≠ Endorsement.

Joined August 2007
Photo by Tramch (@Tramch)Photo by Tramch (@Tramch)Photo by Tramch (@Tramch)Photo by Tramch (@Tramch)Photo by Tramch (@Tramch)Photo by Tramch (@Tramch)Photo by Tramch (@Tramch)Photo by Tramch (@Tramch)Photo by Tramch (@Tramch)Photo by Tramch (@Tramch)
1/ The size of a change has little to do with how it looks on screen. In #ProductDesign, renaming a button is quick. Changing what it does touches logic, layout and every state around it.
1
1
2/ Company language works the same way. A product name in your boilerplate also lives on the website, in the press kit, on the support pages and in the sales deck. That one-word edit is a coordinated release.
1
1
3/ Before anyone calls a change small, list what else reads it. If the list runs past one line, give it an owner and plan it as a project. #Leadership
1
Dario Amodei telling lawmakers that open-source AI is dangerous because you cannot "monitor its usage" or “revoke a user’s access” is the funniest accidental confession in tech history. He literally described why Linux, local encryption, and open weights exist in the first place: so no self-appointed hall monitor can remotely lobotomize your tools or turn off your compute because you hurt their compliance department’s feelings. Calling regulatory capture "AI safety" does not fool anyone who actually builds infrastructure.
Dario Amodei on open source models: “If the path continues, I think we could get to a very dangerous place.” “When you control a model and you're deploying it, you have the ability to monitor its usage.” “You can revoke a user's access. You can change what the model is willing to do.” “When a model is released in an uncontrolled manner, there's no ability to do that. It's entirely out of your hands.” “We should confront the advocates of this with that problem and challenge them to solve it.” He clearly doesn't believe Intelligence should be in the hand of the masses and Sam Altman also saying "RAM" shouldn't be accessible to everyone indefinitely agreed with him.
222
1,548
89
7,730
165,617
Tramch retweeted
Major U.S. banks (and Microsoft) are down.
43
44
16
279
35,552
Tramch retweeted
Robert Pattinson is James Bond
24,529
212,879
84,438
1,377,049
73,064,426
*Offer not valid in all locations*
Your email might be down, but the skies are always open. ✉️✈️
201
2,971
140
35,889
745,645
1/ A synthetic #AI audience can tell you how the average reader reacts. Your message rarely fails with the average reader.
1
36
2/ Research on synthetic respondents keeps finding the same pattern: they match the broad average and miss the variation. Subgroups flatten and strong objections soften.
2
33
3/ Rule: use synthetic panels to find confusion and sharpen your #MarketResearch questions. Before any result changes what you publish, check it with real people from the audience the message is for.
18
Tramch retweeted
Satoshi Nakamoto had an affair?
41
22
4
369
20,231
𝕏 officially expands its photo upload limit, allowing users to share up to 20 photos in a single post
78
214
62
3,911
228,415
Tramch retweeted
Breaking news: Nvidia chief executive Jensen Huang is to depart the company following revelations that he had two extramarital affairs, plunging the chipmaker at the heart of the global AI boom into crisis Tap for live updates ft.trib.al/2OUd1Rg
62
206
48
704
80,139
Brands can now claim their profile inside ChatGPT, while United bars AI agents from booking and the EU wants spending limits on shopping agents. Who owns how your #Brand shows up inside #AI assistants, and which agents you let in?
30
Audience fatigue starts in the content plan. Before a new message gets a slot in your #ContentStrategy, answer one question: what will this audience stop receiving to make room for it? If nobody can say, the plan is adding to the load.
25
I wanted to test Fable 5.2's quality with one-shot prompts without human feedback. Everything in this game is from scratch: models, sounds, textures. ~$550 at API pricing, ~2.7M output and 445M input tokens, heavily cached. Took 8 hours! Play it here: claude.ai/artifact/BGY1SztrA…
857
1,163
846
14,970
4,459,074
Nah, this is diabolical 😭
5
19
2
497
28,991
1/ Your crisis playbook says "align with legal before responding." A colleague calls the duty lawyer. An agent emails the legal inbox at 7am and waits until nine. Agents take playbooks at their word, and most playbooks were written for insiders.
1
53
4/ The risk runs the other way too. An agent follows a bad playbook faster and more consistently than a person would. Each skill file needs an owner, a review date and a change log, the same as your brand guidelines.
1
23
5/ Judgment stays human in #AIinComms: what not to say, when silence beats a statement, who hears first. One test: give a playbook to an agent with a realistic scenario and mark every place it stalls or guesses. Full article: linkedin.com/pulse/your-play…
9