John Paul Helveston

I've been studying China's EV industry for over 15 years. There's plenty of evidence of China's leadership position in EVs, for example my latest article, summarized here:

bsky.app/profile/jhel...

You can also find free access to any of my articles at www.jhelvy.com/research

100% and that's what matters for most Americans. Just one more reason why EV adoption will be much slower here. We won't have electric versions of that for at least another decade, especially since we're keeping some of the best EV tech out of the market and making it less competitive.

A common and quite reasonable reaction to a new product from a new automaker that doesn't have an existing local reputation. It will take a long time to understand that these cars meet or exceed many of the same standards. Of course, when they're blocked from market entry, it will take even longer.

Marques Brownlee on X: "I lived with a Chinese EV for a few weeks to see if the hype is real. The car costs $42K, and turns out it feels like $75K+ EASILY One of the most impressive things I've ever reviewed: https://t.co/EHe358cvLt - A+ software and features. Feels like what would happen if Apple https://t.co/6qBC7m01IT" / X I lived with a Chinese EV for a few weeks to see if the hype is real. The car costs $42K, and turns out it feels like $75K+ EASILY One of the most impressive things I've ever reviewed: https://t.co/EHe358cvLt - A+ software and features. Feels like what would happen if Apple https://t.co/6qBC7m01IT

When I wrote about the technology gap in EVs between China and the US in @science.org, a lot of people emailed me suggesting I was exaggerating.

x.com/i/status/199...

Let's say you have a doctoral student who's a talented programmer but needs an introduction to the foundations of modern ML for a research project. What's the best open source resource for teaching themselves the basics in #python? Assume familiarity with things like pandas.

100% my experience and also why I came here

This is super interesting. I had no idea Haskell had these features, but the arguments are pretty solid for using Haskell for data science

EV recycling is a major issue yet to be fully industrialized, but also one being heavily studied, and processes are being developed. It will take time, and as it gets better the gap in damages between EVs and gas cars will only continue to grow.

The Cost of Pulling Back from China in the EV Transition - Kleinman Center for Energy Policy Energy policy research from the University of Pennsylvania

PODCAST: China now sets the global EV pace. What does U.S. retreat mean for innovation, jobs, and industry leadership? John Helveston (@jhelvy.bsky.social) of George Washington University explains on the Energy Policy Now podcast.

Even if true, tens of thousands would be a rounding error. Chinese OEMs sold 10M EVs in China last year.

The Cost of Pulling Back from China in the EV Transition Energy policy research from the University of Pennsylvania

I had a great conversation with Andy Stone on the Energy Policy Now podcast hosted by @kleinmanenergy.bsky.social on what's at stake for the US as China leads the world transition to EVs

kleinmanenergy.upenn.edu/commentary/p...

NREL is dead, long live...NLOR?

www.nrel.gov/news/detail/...

We're in a new world now for online surveys. Relying on in-person sampling may be the only option for now. At least that way we can monitor the respondent to ensure it's a real person answering the questions without the help of LLMs. Sadly, this means higher costs, less representativeness, etc.

💰 The incentive to cheat is pretty big. The cost to have an LLM take a survey is as low as $0.05 per survey, with payouts starting at $1 and up - a major return if you scale that up. There's also the incentive to just skew data for pure political purposes (e.g., polling, etc).

Generative AI Meets Open-Ended Survey Responses: Research Participant Use of AI and Homogenization - Simone Zhang, Janet Xu, AJ Alvero, 2025 The growing popularity of generative artificial intelligence (AI) tools presents new challenges for data quality in online surveys and experiments. This study e...

Bots aren't the only problem. This paper by Simone Zhang found 34% of Prolific respondents reported using LLMs to help answer open-ended questions. This means that even if we know respondents are humans, we still can't know if the responses are't LLM-generated

journals.sagepub.com/doi/abs/10.1...

The Bots Are Coming for Your Surveys (And They're Smart!) Why commercial panel vendors are sleepwalking into an existential crisis—and what we can do about it

This article by Constantine P. summarizes the PNAS paper (and general validity crisis) quite well if you'd rather read a well-written summary instead of the entire PNAS article.

www.thevoiceofuser.com/the-bots-are...

The potential existential threat of large language models to online survey research | PNAS The advancement of large language models poses a severe, potentially existential threat to online survey research, a fundamental tool for data coll...

A recent publication in @pnas.org by Sean J. Westwood at Dartmouth shows that LLMs can now complete surveys with a level of sophistication renders every quality check useless. The bot in this experiment reached a 99.8% pass rate on attention checks 😮

www.pnas.org/doi/full/10....

YouTube video by John Helveston Comet browser fills out a survey with bot detection questions

Online survey research is screwed. LLM-enabled agents can take surveys just like humans.

Here's a video I made where the Comet browser agent takes a simple survey and gets around nearly all my bot detection checks.

The crisis is here already...more in this 🧵

www.youtube.com/watch?v=05Je...

Yeah this is pretty accurate now. I've had ups & downs using LLMs for #rstats package development, but with Claude code in Positron it's just gotten amazing over the past year. At this point if I don't have an internet connection I won't even start a task because I know I'll solve it faster with AI

Announcing surveydown v1.0.0! – surveydown.org We are thrilled to announce the v1.0.0 release of surveydown, a stable release with improvements to YAML settings, page navigation, question creation, and a smoother overall developer experience.

🎉 Announcing surveydown v1.0.0!

surveydown is an #opensource #rstats package for building surveys with #Quarto & #Shiny. This release improves the survey design experience, including:

- Simplified YAML settings
- Auto page navigation
- Previous buttons!

More at surveydown.org/blog/2025-11...

The app uses web assembly to run everything in the browser, so the page size is considerable. Best run on a desktop and chrome-based browser.

Wright overhauls DOE, reflecting shift in US energy priorities The Energy secretary has reshaped the department to promote fossil fuels, nuclear power and critical minerals.

Definitely a great way to exude more seriousness is to change the name from "Loan Program Office" to "Energy Dominance Finance" I mean for sure it's just so super serious

www.eenews.net/articles/wri...

More in this related thread on a recent piece I wrote

bsky.app/profile/jhel...

Affordability is still one of the biggest barriers to EV adoption in the US. A Chinese presence wouldn't just increase EV sales - it would put the pressure on all US automakers to get costs down. Learning and knowledge exchange from leading Chinese OEMs would also help.

Electrification of light-duty vehicle fleet alone will not meet mitigation targets - Nature Climate Change Electric vehicles (EV) are often considered as the best chance for reducing light-duty transport emissions. Analysis of US policies shows that required emission reductions exceed feasible EV deploymen...

The transition to EVs is critical for reducing emissions, but it's still not enough no matter what the path. Great article on it here: www.nature.com/articles/s41...

If Chinese firms were allowed to sell in the US, my prior is it would accelerate the US transition to EVs

This will destroy US Science

– surveydown.org

Docs on the platform at surveydown.org

YouTube video by Posit PBC surveydown (John Paul Helveston, GWU) | posit::conf(2025)

My talk from #positconf 2025 is now up! I got to introduce our survey platform and #rstats package #surveydown. It's been a fun project with @pingfanhu.bsky.social & Bogdan Bunea, and I hope it can be useful for the community of survey researchers out there!

www.youtube.com/watch?v=Vwoe...

dplyr but make it bussin fr fr no cap `genzplyr` is an alternative syntax for `dplyr` that replaces boring old function names with GenZ slang. Your data wrangling is about to hit different.

Thanks to this bussin new #rstats 📦 by @hadley.nz,
you can now yeet rows and vibe_check columns in your data frames, no cap

hadley.github.io/genzplyr/

Some noted that the BEV ranges in my article were unadjusted (China's CLTC cycle over-estimates range compared to the U.S.'s EPA cycle). So I now have an interactive app that allows you to explore how the main scatterplot looks with an adjustment factor:

jhelvy.github.io/science-2025/