Tbh I assume that everyone using LLMs in prod is building a failover system. The team wrote about ours blog.gradient-labs.ai/p/building-r...

PSA: go apply the urgent security update to your Apple devices. Don’t wait. Do it right now

We wrote a post about a couple inter related incidents we had a few months ago blog.gradient-labs.ai/p/anatomy-of...

Any European startups need this data? It’s about to go up in smoke.

A new definition for AGI just dropped, and it is a bad one.

Monzo alumni's Gradient Labs raises $13M to reinvent agentic AI for regulated industries tech.eu/2025/07/08/e...

Gradient Labs secures $13M Series A to revolutionize AI-powered customer service in regulated industries, delivering 90% query resolution with 98% quality assurance #AI #CustomerService

💰 Series A! gradient-labs.ai/blog/series-...

AI is not going to make people atrophy into bad writers. People are already bad writers

After decades, I still hear the London tube thing as “see it, say it, sort it”

The best thing about today’s outage was every b2b company announcing to each other that they’re impacted

love an outage day

We need a 3D Designer/Animator at Sling for a few weeks, if you know anyone send them this listing: app.screenloop.com/careers/avia...

An announcement: After 18 years, TechCrunch and I have parted ways. www.linkedin.com/posts/mikebu...

The state of marketing for b2b AI agents in 2025

Am currently looking to hire Backend & Product Engineers. Not sure if shouting into the void here will help, but if you - reader - are or know someone great, please let me know!

This is a great read 👇

To all the people who said a few years ago that agents like this were coming, props for the forecast. These agentic LMs are such a crazy platform to build a ton of new apps and tools on! So much capabilities untapped in recent models.

🎧 I got invited to the “Between Data & Risk” podcast Internal team discussion: should I have (not) shaved for this? youtu.be/xMJ06jC51VY

Hopefully things change here, as my big worry about AI talk on this site is that people are being misled about the capabilities of AI by influencers who are telling them what they want to hear. AI is not going away, it will impact all of our futures, and it needs more voices shaping its direction.

Most important plot from IO today -- AI usage is skyrocketing. This is real.

We’re so close to real magic

Saw a post on LI claiming there is no credible peer-reviewed evidence of a real world agentic AI deployment (according to their definition of agentic) and that everything is BS. So many layers. I got stuck trying to think of when peer-review became the standard of evidence for real deployments

What a horror show to have put that man in charge of anything. The documentary is very short and very good. You can watch it here: www.youtube.com/watch?v=ybVZ...

A common question is "can an AI make money?" This benchmark, where AIs run a simulated vending machine over time, suggests yes, with an important caveat On average, Claude 3.5 & o3-mini beat a human, but are high in variance & fail at random times for complex reasons. andonlabs.com/evals/vendin...

Awesome thread on language and llms

genuinely one of the coolest tech presentations you can find - Google evolution 1999->2009. static.googleusercontent.com/media/resear...

Approved a fix fora time zone issue today so had to dig up this NormConf gem youtu.be/D48NQTNg19s?...

I wonder what the “if we multiply numbers enough it’ll happen” crew thinks of this www.scientificamerican.com/article/wher...

Redis is open source again: antirez.com/news/151

If you heard the news about Pepperius, then you’re in my clan 🐷

"If you want to prompt ChatGPT 40 times, you can just stop your shower 1 second early." "If I choose not to take a flight to Europe, I save 3,500,000 ChatGPT searches. this is like stopping more than 7 people from searching ChatGPT for their entire lives."

Decades in, careers established, job titles refreshed and we still ultimately split into “day-ta” and “dah-ta” people

The killer app for AI is turning meetings that should have been an email into an email.

We’ve been using Graphite for their stacked PRs for a while now and they send me a weekly run down. Majority of our changes are small, tightly scoped and quickly shipped 🤩 (no vibe coding for us!)

This is one of the worst violations of research ethics I've ever seen. Manipulating people in online communities using deception, without consent, is not "low risk" and, as evidenced by the discourse in this Reddit post, resulted in harm. Great thread from Sarah, and I have additional thoughts. 🧵

😂😂😂😂😂 velvetshark.com/ai-company-l...

how the email found me:

This keeps coming up, so if you were wondering, we rigorously tested whether being polite to AI matters & it seems to have no effect on answer quality in aggregate. It increases the quality of particular answers while lowering the quality of others, and it is not possible to know which in advance.

Imagine how many GPUs they could get on a distributed training run if ... it was a game people could play on their home console app.primeintellect.ai/intelligence...

I used to think that a plain text website on blank background meant “I’m a CS professor,” but now it has also come to mean “we’re a futuristic AI startup”

Whoever invented the packaged triangle sandwich doesn’t know the profound effect they had on the culture of this country

In answer to “which LLM do you use at Gradient Labs?” … new blog post: blog.gradient-labs.ai/p/llms-at-gr...

This is so incredibly cool. Embeddings are lingua franca for so many retrieval/LLM/recommendations use-cases and I really enjoyed working with embeddings in Redis via the search module, can’t wait to try this.

At Gradient Labs we’re celebrating our first £1M ARR 🎉 My cofounder got interviewed about it here: www.growthmentor.com/location/lon...

What do real careers in AI look like? Hear from @neallathia.bsky.social and his journey to co-founding an AI startup 👇 medium.com/@catherinebr...

“We’re still ablating the optimal mixture of 4chan tokens in our post-training set”

🧵 1/ Me and a friend have developed a lil' agent you can ask research questions to, it will use the arXiv API to find relevant papers and respond. It can also summarise recent developments in an area. It's a beta for now, looking for feedback if anyone's keen to do us a favour? shorturl.at/5Pdxq

This Github ticket is the kind of proposal that inexperienced engineers tend to make, because they think it's about code, and not about the humans who use the system. In any competent organization, these junior folks would get some very direct feedback and it would be a growth moment.