@inerati “You keep messing up with curl, just use Python in the shell”
English
Jacob Peddicord
304 posts
“format this list of strings into a python list” thinking mode on, effort max
arb8020@arb8020
i wonder when people will realize that for the majority of use cases using the Biggest Newest GPT5.8 Claude Opus 6.7 for everything kinda looks like this
English
This stacked icon design inspired the look of the "Copy for AI" button we just added to the Exa playground
One click to get the perfect context string for any LLM
Brotzky@brotzky
For your stacked elements: Stop using border or box-shadow. Start using mask.
English
is there a famous domains broker? i feel like i saw something to this effect on twitter. asking for a friend writing a book on domains
English
Might be the craziest demo I’ve seen since ChatGPT. I didn’t know we were so close to this.
Logan Kilpatrick@OfficialLoganK
Introducing Genie 3, the most advanced world simulator ever created, enabled by numerous research breakthroughs. 🤯 Featuring high fidelity visuals, 20-24 fps, prompting on the go, world memory, and more.
English
Jacob Peddicord retweetledi
@blakeir The point is to get objective data!
Why are so many people afraid it will harm their subjective experience.
English
sleep-tracking apps should ask “how rested do you feel?” before they spit out a score.
English
@nickarner In the Twitter slop stage, happy to be proved wrong
English
Is anyone actually using mcp servers day-to-day, or are all these in the toy / demo stage still?
English
@atroyn Is the market that inefficient? Or are the returns just not high enough?
English
@jacobpedd to pull it forward faster than the market would ordinarily build it.
English
what the fuck are we doing here
Michael Thomas@curious_founder
The US will need ~450 GW of new electricity generation capacity by 2030 to meet rising demand from data centers, reshoring, etc. And the Senate is about to vote on a bill that could wipe out ~500 GW of potential energy generation capacity.
English
@atroyn Why do we need subsidies if there’s going to be market demand? Serious question
English
“wind and solar need subsidies to be viable” yeah those subsidies are covered by the productivity growth of building the power! we’re just pulling it forward faster
anton 🇺🇸@atroyn
@WatcherontheWeb hey what if we brought as much electricity as possible online as fast as possible and the subsidies canceled out in two years because of the productivity growth it would fuel. we’re just pulling power forward and we need to.
English
Life update, I joined @ExaAILabs as a TPM to help grow their search API.
If your AI product needs to talk to the internet, we should chat!
English
Jacob Peddicord retweetledi
Introducing Exa Research - agentic search that doesn’t stop until it finds what you need.
Research scores 94.9% on SimpleQA, the highest of any research API. We also do it at the lowest price, by using cheaper models for easier tasks, smarter models for harder tasks.
The endpoint is especially good at structured outputs, but can also be used to generate in-depth reports (similar to Deep Research products).
Try it in our API 👇
English
@blakeir ~Silent~ voice input would be even better
Voice has like 2x the throughput of typing
There's just no good hardware solution rn
English
the ideal AI product experience is:
- voice input
- text output
English
Jacob Peddicord retweetledi
1/ We’re humbled to announce @southpkcommons Fund III: $275M to support exceptional founders from day -1.
Since 2016, we’ve had a simple thesis: greatness is more likely to emerge when high talent density meets high curiosity. That's why we focus on -1 to 0.
English
I had a lot of fun hacking over the weekend with @stehl_c @robechun @delaho_h. They're all crazy talented and made coding until sunrise fun.
Felt great to get the win!
Shout out to SPC for organizing and loved seeing what everyone built.
Finn Meeks@finn_meeks
🏆SPC x Anthropic Hackathon GRAND PRIZE: Nardwuar–the on-demand researcher and report generator for all of your interviews (Nardwuar level research, without the Nardwuar level time). Built by @stehl_c, @jacobpedd, Robert Chung, & @delaho_h
English
Jacob Peddicord retweetledi
this was lots of fun and lots of all nighters over the past few weeks. really happy with what we achieved!
Artificial Analysis@ArtificialAnlys
NVIDIA Blackwell can achieve 303 output tokens/s for DeepSeek R1 in FP4 precision, per our benchmarking of an Avian API endpoint Artificial Analysis benchmarked DeepSeek R1 on an @avian_io private API endpoint. Running DeepSeek R1 in FP4 precision on NVIDIA Blackwell, their endpoint achieved 303 output tokens/s - the fastest speed we have measured yet for DeepSeek R1. The FP4 version of DeepSeek R1 maintained accuracy across our evaluation suite as compared to the native FP8 version. Inference speed is especially critical for reasoning models that ‘think’ before they answer - we look forward to wider availability of NVIDIA Blackwell hardware in the coming months!
English