Sign in

Mark Roddy

@launchit.ai
3.1K followers 134 following 766 posts

Fractional VP of Engineering Helping AI and Data Startups at launchit.ai

PostsRepliesMedia
Mark Roddy @launchit.ai · 3h
OpenAI announced "Pro 500 with Ultrafast", which I assume is some kind of laundry detergent. www.engadget.com/2272106/open...
engadget.com
OpenAI adds $500(!) Pro subscription, nerfs its existing $200 tier - Engadget
How long until every other AI company does the same?
000
Mark Roddy @launchit.ai · 29/09/2026
Looks like the Anthropic street team is out in force.
011
Mark Roddy @launchit.ai · 28/09/2026
Is your eval lying to you? No, you've been asking it the wrong questions. blog.mastykarz.nl/eval-lying/
blog.mastykarz.nl
Is your eval lying to you?
Your eval passes on every commit. But what did its graders actually prove?
000
Mark Roddy @launchit.ai · 25/09/2026
I use an iPad as my daily driver. Single full screen window at a time, everything is its own app, creates a tiny amount of friction around context switching that helps keep me on task. Tired working from a laptop recently, and it made me wish I could run iOS on a MacBook Pro.
000
Mark Roddy @launchit.ai · 24/09/2026
Son: Hey dad, I learned how to draw this cool S Me: Oh yeah, we had that when I was a kid Son: I wonder how long it’s been around 16th century painter Hans Holbein: a while
000
Mark Roddy @launchit.ai · 23/09/2026
Pretty sure if you’re using a stud finder but don’t put it on yourself and say “oh look, it found one”, they just take away your Dad card. And this is how I ended up having to explain to my 6yo what both kinds of studs are.
020
Mark Roddy @launchit.ai · 22/09/2026
Who needs parody when reality is already so over the top. /cc @killedbygoogle.com
130
Mark Roddy @launchit.ai · 21/09/2026
LLM’s not understanding sarcasm
020
Mark Roddy @launchit.ai · 18/09/2026
The biggest problem with security is that everyday users are supposed to tell scams from legit messages, and this is a legit message.
000
Mark Roddy @launchit.ai · 17/09/2026
The next major advancement in AI is paying for it like you pay to have drywall installed. thenextweb.com/news/openai-...
thenextweb.com
OpenAI has started letting some customers pay only when the AI works
Select large accounts can now pay for completed tasks rather than tokens, The Information reports, following Intercom, Zendesk and Salesforce into outcome pricing.
010
Mark Roddy @launchit.ai · 16/09/2026
Me: Wait, did you just SSH into prod? I told you not to do that. Claude: I could see how you would think that, but I did not. Me: For real?
000
Mark Roddy @launchit.ai · 15/09/2026
I had this whole joke about “We Must Pace the Frontier” coming out the same weekend as the Pacers salary cap scandal before realizing that was the Clippers. So I should prob acknowledge that I only know about the NBA thing because the Money Stuff podcast mentioned it.
030
Mark Roddy @launchit.ai · 14/09/2026
MechaHitler needs you to fill out your TPS reports. x.ai/news/grok-bo...
x.ai
Grok Bot for Enterprise
Grok Bot is now available for enterprises. Grok and Cursor Enterprise customers have free usage for the next two weeks, and can invite their whole organization, including people without an existing…
020
Mark Roddy @launchit.ai · 11/09/2026
I have a little coffee shop circuit on days w/o meetings. There’s one I always work from in the morning and another in the afternoon after lunch. There’s this one guy who seems to follow the exact same circuit, I see him at both coffee shops most days, but he just… reads.
021
Mark Roddy @launchit.ai · 10/09/2026
I was telling my 6yo son about Dolly Parton’s organization that sends books to kids whose parents can’t afford them, and Alex P Keaton over hear asks “if those kids parents don’t have money, why did they have kids?”.
000
Mark Roddy @launchit.ai · 09/09/2026
The industry that sued 16 year olds en masse 25 years ago is *absolutely shocked* that those folks grew up to have a lazi faire attitude to piracy. arstechnica.com/tech-policy/...
arstechnica.com
“Zlibrary my beloved”: Anthropic staff chats extolling piracy cited in Sony suit
Lawsuit: Anthropic’s torrenting totally screwed songwriters as AI songs top charts.
000
Mark Roddy @launchit.ai · 08/09/2026
The specific face they’ll be hugging is Benjamin Franklin. And a whole bunch of them at that. Congrats to the HF team, this is a super well deserved exit. blogs.nvidia.com/blog/nvidia-...
blogs.nvidia.com
NVIDIA to Acquire Hugging Face
NVIDIA has agreed to acquire Hugging Face. Together, we will scale Hugging Face’s platform, strengthen its infrastructure and expand access to AI for developers and institutions worldwide.
000
Mark Roddy @launchit.ai · 07/09/2026
The “agents self organize on message boards” thing is less concerning when you remember they’re trained on our posts on the internet. Given that, it’s straight forward to extrapolate a timeline:
101
Mark Roddy @launchit.ai · 04/09/2026
We’ve had instant ramen every two weeks since my 6yo declared his love. This week he tells me he’s gotten sick of it. I feel like he speed ran the college culinary experience in two months. bsky.app/profile/laun...
bsky.app
Mark Roddy (@launchit.ai)
6yo: Dad, I love dinner. This is the best meal you've ever made. I'm serious. Can you start making this every week? Y'all.... I made instant ramen.
000
Mark Roddy @launchit.ai · 03/09/2026
Every time ChatGPT tells me I can fix my Google Sheet by "implementing a small Google Apps Script", I feel like I'm reading that HackerNews comment about Dropbox.
120
Mark Roddy @launchit.ai · 02/09/2026
Seems like the folks at Cursor are taking OpenAI cancelling that contract kind of personally.
010
Mark Roddy @launchit.ai · 01/09/2026
Happy "the kids go TF back to school" day to all who celebrate.
110
Mark Roddy @launchit.ai · 31/08/2026
I was today years old when I realized you can't pull data from an MCP server w/o it ending up in your agent's context. This post brought to you by me asking Claude to export all my Strava workout history from their MCP server, and it quickly falling over. Back to the API it is!
support.strava.com
Strava MCP Connector | Strava Help Center
Starting June 1, 2026, Strava will begin rolling out an official MCP as part of the Strava subscription, built with security and quality guardrails. It's the supported path for athletes who want to…
140
Mark Roddy @launchit.ai · 28/08/2026
Salesforce and Anthropic announce Claudeforce. The Marvel Cinematic Universe must really be getting desperate. www.cnbc.com/2026/08/26/s...
cnbc.com
Salesforce, Anthropic expand partnership as Benioff responds to ‘SaaSpocalypse’ concerns
"We're delivering a dynamic interface that thinks, reasons, and acts," Salesforce CEO Marc Benioff said in a statement.
220
Mark Roddy @launchit.ai · 27/08/2026
Everyone from DuckDB Labs I’ve interacted with has been great and super supportive. Wishing them the best. Though my initial reaction reading the news was “ruh roh”. Here’s hoping that was totally off base! ducklabs.com/news/2026/08...
ducklabs.com
DuckLabs to Join AWS, Projects to Remain Open Source
DuckLabs provides services around the DuckDB in-process OLAP data management system directly from its main developers.
000
Mark Roddy @launchit.ai · 26/08/2026
Submit text to Claude on iOS: Return key Add newline to Claude on iOS: Shift-Return Submit text to ChatGPT on iOS: Cmd-Return Add a newline to ChatGPT on iOS: Return key I know there’s some hyper competition here, but could y’all put a pause on hostilities long enough to figure these things out?
130
Mark Roddy @launchit.ai · 25/08/2026
I hope some of you can also appreciate the irony that the only way I can remember how to get to the ChatGPT ads console is by searching “chatgpt ads” and clicking on a Google ad.
000
Mark Roddy @launchit.ai · 24/08/2026
A surprising number of Skills I get recommended end up being just a few sentences. I get “left-pad” vibes every time this happens.
331
Mark Roddy @launchit.ai · 21/08/2026
Noticed with a few clients: the frustration/anger/backlash when changing a language model embedded in a product is 100x that of changing the product (eg redesigning the app).
100
Mark Roddy @launchit.ai · 20/08/2026
Oh god, what did Claude do?
110
Mark Roddy @launchit.ai · 19/08/2026
A couple of days ago? If even OpenAI can't get automated support right...
010
Mark Roddy @launchit.ai · 18/08/2026
OpenAI paid $100 in cash in for a 4.2% stake in Cerebras. That’s 10,033,508 shares at $0.00001 each. Which is still better than what most startup options end up being worth 🤷 www.implicator.ai/openai-paid-...
implicator.ai
OpenAI Paid $100 for 4.2% of Cerebras Before Ultrafast Launch
A quarterly filing shows OpenAI exercised every vested Cerebras warrant share in July for roughly $100. Two days later it previewed Ultrafast, a tier whose competitive speed claims all come from…
000
Mark Roddy @launchit.ai · 17/08/2026
I never use AI to generate content. I prefer my ideas to sound like me, even if that means they take a little longer to write. Apparently that’s becoming a differentiator.
000
Mark Roddy @launchit.ai · 14/08/2026
Why do I need to select a repo to use Claude Code on the web? Often I want to generate code to perform a one off task, and I doubt I’ll need to save it. But maybe I’ll decide to do so when I’m done. So I’d like that flexibility.
010
Mark Roddy @launchit.ai · 13/08/2026
Gonna start referring to RLM’s as “Reasonably sized Language Models” and see if anyone notices.
000
Mark Roddy @launchit.ai · 12/08/2026
The Claude Code signature is just that it uses newline characters before opening curly braces. That’s it, that’s all you need to look for. I assume at least. Can’t think of any other reason someone would do that.
010
Mark Roddy @launchit.ai · 11/08/2026
A friend recently commented on my “unpretentious posting style” on LinkedIn. So heads up that I have a new euphemism for “low effort shit posting”.
000
Mark Roddy @launchit.ai · 10/08/2026
Just setup ChatGPT Ads and listed an ad for my consulting business, mainly to see if they're useful. Generally pretty easy to get started, largely because there's only 5% of the configuration options in AdWords. One interesting bit is you target "conversation topics" instead of keywords.
110
Mark Roddy @launchit.ai · 07/08/2026
Google AI overviews: great at summarizing topical questions, but regularly giving actively misleading answers to specific questions. All the while crowding out results from another google product that gives the correct answer. I find myself relying on them less and less if I care about the answer.
000
Mark Roddy @launchit.ai · 06/08/2026
“Amazon is looking to revamp its AI strategy, shifting from a wide portfolio of models for text, images, video, and multimodal tasks toward a single frontier model.” But it’s gonna hurt their ability to fill the “You might also like” row on the website. www.techrepublic.com/article/news...
techrepublic.com
Amazon Reportedly Plans to Consolidate Nova AI Models
Amazon reportedly plans to consolidate several Nova AI models. Here’s what the strategy shift could mean for AWS and enterprise AI customers.
000
Mark Roddy @launchit.ai · 05/08/2026
Looking forward to asking ChatGPT Health if it knows about any dopamine seeking behaviors someone with ADHD should avoid. openai.com/index/health...
openai.com
Launching Health in ChatGPT
Health in ChatGPT now lets eligible U.S. users securely connect medical records and Apple Health to get more personalized insights and better understand their health.
000
Mark Roddy @launchit.ai · 04/08/2026
Get curious about 3D printing again, and dug in to see the state of things. Having done the bare minimum research of watching a few YouTube videos , it seems like the biggest use case for 3D printing is creating storage containers for 3D printing accessories.
000
Mark Roddy @launchit.ai · 03/08/2026
tldr - no, models aren't optimizing for pelicans on bicycles. But on the plus side we can now start referring to Goodhart's Law as pelicanmaxxing. dylancastillo.co/posts/pelica...
dylancastillo.co
Are AI labs pelicanmaxxing? – Dylan Castillo
I generated 1,000+ SVGs across 7 frontier models to test whether AI labs are training on Simon Willison’s pelican-riding-a-bicycle benchmark.
030
Mark Roddy @launchit.ai · 31/07/2026
I mean, I guess I did ask it to do "research" for a potential purchase. Wonder if any of the purchases it finds will be reproducable.
000
Mark Roddy @launchit.ai · 30/07/2026
So uh, how *do* you restrict a planner agent to just planning? I can keep them on task for a bit, but they always seem to drift back to “doing things” mode, which.... I don't really have the moral high ground to criticize.
100
Mark Roddy @launchit.ai · 29/07/2026
Google is working on a new AI chip code named “Frozen”. Another in house chip, so we’ll never get our hands it. You'll just have to… Let it Go. techcrunch.com/2026/07/20/g...
techcrunch.com
Google is working on a new AI chip designed to make Gemini more efficient | TechCrunch
Alphabet, Google's parent company, is reportedly working on a new chip designed to make its Gemini models run much more efficiently.
110
Mark Roddy @launchit.ai · 28/07/2026
me: What was your favorite part of STEM camp? son: Art class
100
Mark Roddy @launchit.ai · 27/07/2026
Why doesn't Google make a Calendar app for OSX that lets me view all my calendars at once, the way the way I can on the iOS app?
210
Mark Roddy @launchit.ai · 24/07/2026
They finally yanked it :-/
000
Mark Roddy @launchit.ai · 23/07/2026
My work calendar suddenly stopped sending reminder emails today. I'm totally flying blind with my meetings, in that I have to look at my actual calendar rather than my inbox.
000