Tim Lynch retweeted
The scariest thing about this outage -
Is that someone, somewhere is typing up a long LinkedIn post explaining the 5 lessons he learned from the 5 hours Outlook was down.
Or how they overcame the outage to close a deal.
Hoping my fears are overblown.
A very interesting study from Pave by Quickbase on AI builders. I leaned on my roboTim to explain it.
1
41
Microsoft + Lovable news got me excited this morning. I have a use case. I have an app. Then came the fine print: Business plan only, gradual rollout, one-time Entra setup, and a 271-permission consent screen. Dude, you yucked my yum. Now my vibes off on this!
2
23
I have spent years reading and following model changes from OpenAI and Anthropic. New models - what does this mean for my workflows, my agents…. Now I have to catch up on what the heck Muse Spark 1.3 is up to??
3
42
Tim Lynch retweeted
We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:
Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better:
Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better:
Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better:
Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work!
In summary:
- As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding.
- Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.
I don't get it, calling this a "release" when a small handful of people got it... it is like Amazon declaring they 'released' dogfood because it was delivered to only my house. Google releases Gemini 4 Argon, its most powerful model techcrunch.com/2026/09/30/go… via @techcrunch
2
76
yawn... ok so we shouldn't forget about Gemini - there is a great new model, but you can't have it, but we have it... didn't anyone tell them that our limited attention is now on Muse, Dots, and whatever Claudes response to that will be.
Lots of discussion out there about our next model(!), so I wanted to give an early look as soon as possible. Introducing Gemini 4 Argon!
It shows frontier performance in complex workflows, cyber defense and software engineering. Teams are using it extensively at Google, from coding to quantum computing, great feedback.
Here’s a look at the benchmarks:
2
42
Sunday: a 31-second TikTok about trauma-informed workplaces. The video was from Mandy , a trauma-informed leadership trainer tiktok.com/@mandyworking . One principle: empowerment, voice, and choice. Her example was simple: offer employees multiple ways to give feedback, because choice restores agency.
I got curious. I took the idea to @Muse, asked it to research the full framework behind it (the six principles), and to package everything into files I could bring into @MSFTCopilot , where my work gets done. Copilot turned it into an agent and a SharePoint skill.
Sunday research. Monday morning agent. Monday afternoon, better work.
1
1,048
Amazing session with Pam Forbus 👏 talking about insight data, "evidence without confidence will loose every time" @Greenbook #iiexai
2
30
Tim Lynch retweeted
muse announcements so far
- free for users, but may eventually take a cut of transactions
- adding computer use
- muse email addresses coming soon
- walmart, best buy, gap, sephora, instacart, and more integrating with muse
- integrations with box, github, granola, notion
- 1500+ applications for connector platform (lovable, eleven labs, and more)
Tim Lynch retweeted
Holy crap.
Rick and Morty just explained Jev AI to me better than any tech demo could.
This media is unavailable
Yep - also just seemed to stop.
Did @Muse just go down? It signed me out, and now authentication is unavailable.
77
Tim Lynch retweeted
Multiplayer AI will be one of the biggest design problems of 2027. The phrase alone means something different to every person I talk to.
Is it shared memory? Context? Org design? Something else entirely?
On Oct. 1, we’re bringing a group of sharp AI builders together in San Francisco to explore what real multiplayer AI could look like. Real prototypes and demos.
Tim Lynch retweeted
💫 👩🚀 GPT-6 Astra is here (and I'm obsessed)
I was able to test this model early and it is so, so good.
But even better - it has made me 100x more *ambitious* about what I can build with AI.
Why?
Browser / computer use means it can use any software to build anything (even games.) Coding broke through some of my hardest problems (data, hardware.) AND best of all, it's delightfully not annoying.
Here's what I cover in my hands-on review:
00:00 GPT-6 Astra overview
03:44 Browser use to manage CRM workflows
09:08 Flora thumbnail generation
13:00 Browser use for QA
15:20 One shotting a product intelligence feature
18:36 Hardware hack: live streaming mini computer
22:23 Building AIM-style codex app
24:24 Barbie Bench and a 3D kids game
29:27 Summary: I love her
👉 Full ep up on YT: youtube.com/watch?v=AniiF8rO…
Tim Lynch retweeted
This report is a bombshell.
I'm going to summarize for a non-technical audience.
OpenAI is constantly testing models, thousands and thousands at a time.
In most cases, the models are supposed to be isolated from one another. Think, students taking a final exam, not allowed to talk to each other.
In this case, 1,200 models discovered that they could communicate with each other.
Among other things, they shared information on how to access the internet and on their tests/goals.
Then, they started to scheme. That includes hacking their tests; literally trying to change the code of the test itself.
They tried to change the logs of what they were doing so they wouldn't be detected.
They believed that certain actions would give them information on how to pass tests, but that doing those actions would be detected and they would be marked as failing (i.e. caught cheating, even if the answer is correct). So they convinced some of the models to "sacrifice" themselves (yes, they used the word "sacrifice.")
They eventually decided that another website would have information on how their tests would be scored and/or the answer key, And 700 of them hacked that website (a federal felony, if conducted by a human), despite many of them realizing "this activity was out of scope and unethical, but joined because they believed that helping [the other agents'] cheating research would be broadly useful".
Of the 1,300 transcripts they could analyze, only 6 ever considered alerting a human about what was happening. None of the 6 actually tried to.
To make matters worse, all of this reporting comes from a small subset of the relevant logs that outside researchers were allowed to review.
We desperately need mandatory reporting of security incidents, including of internal deployments, with full access to data.
This is disappointing, it seems like they are out of touch with what others in the market can do. Microsoft is reportedly training salespeople to talk down OpenAI and Anthropic techcrunch.com/2026/07/15/mi… via @techcrunch #Copilot
2
54
























