frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

ChatGPT lost 22 points of web share in a year

https://aicharts.grok.me/c/market-share
17•echohive42•1h ago

Comments

echohive42•1h ago
Worldwide AI-chatbot web visits: ChatGPT 76% → 54%. Gemini 6% → 28%. Claude 1% → 9%. This is not monthly users and not revenue.
andai•45m ago
I understand Claude being niche (until they started running Superbowl ads?), but why'd it take Google this long to start catching up?
fzzzy•43m ago
Because Google's AIs are absolutely awful?
bdangubic•41m ago
they are good enough for 89.56% of ChatGPT users
andxor•41m ago
Really? This just tells me you haven't tried it in a long time. I'd argue it's a much better experience for the average user.
a2ff6eeb0•40m ago
Because they were slow rolling them out, and heavily focused on the profitable corporate customers.
Forgeties79•50m ago
Alternate headline:

“After three years of unprecedented investment including governments, Facebook, and Google dumping money into the promise of AI, ChatGPT still has over 50% of the traffic.”

Also, grok isn’t been on that list, further proving that no serious people use it.

A_D_E_P_T•20m ago
Grok is there, just a little bit too small to label. Dark brown on the chart.

4.6 is a good model, by the way. Legitimately frontier-level and a good communicator.

PufPufPuf•14m ago
Whose worldview is it communicating, though?
xyzzy_plugh•46m ago
Gemini is actually good enough for pretty much everyone I know who is not in tech and uses Android. These people are downright pleased with Google's AI summaries.

The fact that Google can push Gemini out in so many ways with their existing marketshare is mind boggling. The fact that ChatGPT hasn't lost 50-100% of web share shows it's Google's game to lose at this point.

ipdashc•37m ago
> These people are downright pleased with Google's AI summaries.

Consider me part of "those people", I guess, but are you guys not, at this point? There was a period near the start where they were bad, but I find them perfectly usable now. I'll often open a tab, Google something, and just read the summary because it's basically instant and I don't need any more than just that, a summary.

I'm genuinely confused why people take so much issue with them. Sure they're still wrong sometimes, but it's blatantly obvious when that's the case, and I can just scroll past.

paimapi•33m ago
have you ever done the fun trick of asking any model to adversarially review itself? highly recommended practice for anyone who trusts a one-shot output
ipdashc•22m ago
Sure, I've done this for more complicated questions and sometimes it reveals flaws. But I don't use Google for anything complicated. Generally it's just to get a quick vague definition of a term, pretty much.
Retr0id•32m ago
They're ok. But when I use a search engine it's because I want search results. If I wanted an LLM's interpretation I'd have just asked a search-enabled LLM in the first place.
thot_experiment•44m ago
54% to go, good work team.
razster•32m ago
Meme: "I'm doing my part" Starship Troopers.
edoceo•21m ago
Service guarantees citizenship!
dyauspitr•10m ago
Why? Who is the replacement? Some state backed Chinese thing that is running loss leaders for a rug pull in the near future like Deepseek just did?
Retr0id•43m ago
I think it'd be more informative to see total users per provider, rather than % shares.
AndrewKemendo•41m ago
Given this anyone still thinking Google lost needs to update their belief system
fragmede•34m ago
Lost at what, though? Waymo is gonna pan out but they sold their domain business to squarespace despite having GCP, and Google workspace deletes data when changing billing which is absolutely unacceptable.
xyzzy_plugh•21m ago
The domain business is a distraction. It was wise to sell it. They still own several gTLDs so they have plenty of revenue from this avenue.

Selling .coms and Donut's gTLDs? Peanuts.

Godaddy makes all their money from other stuff, and Google still does lots of other stuff. Google Workspace still prints money.

0xbadcafebee•37m ago
I used to use Gemini (the product) because it's free. But its responses are so incredibly bad, and the interface has so many dumb bugs, that I stopped using it and switched to OpenCode Go models + Web UI. Gemini models can recite deep knowledge, but they can't do constructive things, can't do extended web research, can't provide more than a couple options for an answer. Either their system prompt is nerfed like crazy or the models just suck at agentic work. Even ChatGPT from 2025 was better than Gemini today.
inerte•10m ago
What's the methodology? Is Gemini actual gemini.google.com or mixed with google.com and Android searches?
ipdashc•24m ago
True, but I find it's significantly faster than even "instant" chatgpt, and comes with the advantage that if the summary isn't what I need, I can just scroll down to see the real results (granted, which are usually useless these days).

I agree they should make them disable-able though.

dpark•13m ago
> If I wanted an LLM's interpretation I'd have just asked a search-enabled LLM in the first place.

When you search on Google now that’s exactly what you’ve done.

Retr0id•11m ago
Right, that's the problem I'm describing.
stingraycharles•36m ago
If only they could make it be wrong less? I know they need to have these results appear in record time, but I wonder the collective damage being done by just how horribly wrong the answer often is.
edoceo•22m ago
Can you point an example? For me, recently, it had the wrong price for cable-car fare in SF, but it did show (2015) in the result. That seems reasonable to me, it's not horribly wrong. Other stuff it's been accurate on.
razster•33m ago
I've been using Gemini as an advance search engine. Their main search with AI isn't that useful but Gemini is. They're the only AI online I will use and just for searching/compiling. For work/development it is Qwen3.8. Then again now that I found out Pi has a web search function I've been using Gemini less and less.
dyauspitr•13m ago
But are they paying for Gemini?
dyauspitr•12m ago
It’s good. Returns a bespoke response is some ridiculous 0.3s or something like that which is mind blowing and I have no idea how they are pulling that off at all in the first place let alone the sheer scale of it.

Asus Bike Booster

https://www.asus.com/accessories/bike-booster/asus-oxiis/oxiis-intelligent-bike-booster/
47•wiradikusuma•3d ago•16 comments

Semaglutide linked to lower predicted dementia risk

https://alz-journals.onlinelibrary.wiley.com/doi/10.1002/dad2.70432
347•randycupertino•9h ago•241 comments

Cultivating a state of mind where new ideas are born (2023)

https://www.henrikkarlsson.xyz/p/good-ideas
78•felixbraun•4h ago•20 comments

AI in drug discovery – what it is, where we stand and the path forward

https://www.science.org/content/blog-post/so-how-ai-drug-discovery-doing-really
86•AnodicElegy•6h ago•43 comments

At-home test for infected ticks could improve Lyme Disease diagnosis

https://www.smithsonianmag.com/innovation/the-first-at-home-test-for-infected-ticks-could-improve...
211•gmays•11h ago•77 comments

Abdominal fat predicts heart disease risk better than BMI

https://www.acc.org/about-acc/press-releases/2026/08/11/14/59/abdominal-fat-predicts-heart-diseas...
150•theanonymousone•4h ago•113 comments

Tea5767-Radio-Tuner

https://github.com/turtushig22-blip/tea5767-radio-tuner
11•turtushig22•1h ago•0 comments

RISC-V: They Should Have Known Better

https://dmitry.gr/?r=06.%20Thoughts&proj=12.%20RV
233•dmitrygr•1d ago•298 comments

Super El Niño Keeps Growing as New Forecasts Reach Record Territory Ahead Winter

https://www.severe-weather.eu/long-range-2/super-el-nino-growth-accelerating-to-record-strength-f...
77•dgellow•6h ago•31 comments

Tracking down a Zsh history data loss bug

https://michael.stapelberg.ch/posts/2026-08-09-zsh-history-truncation-bug/
34•ingve•4h ago•6 comments

Auto-research with codex: How I achieved a 232x Faster Kernel

https://sankalp.bearblog.dev/autoresearch/
394•tosh•14h ago•86 comments

AI has access to a vastly larger working memory than the human brain

https://davidepiffer.com/p/ai-isnt-outthinking-mathematicians
412•rzk•7h ago•372 comments

A fortuitous decade as an indie software developer

https://lapcatsoftware.com/articles/2026/8/3.html
28•frizlab•5d ago•2 comments

A spectre is haunting Unicode

https://www.dampfkraft.com/ghost-characters.html
171•sensanaty•11h ago•57 comments

SugarTrack – an offline Android logbook for blood sugar (no account, no cloud)

https://sugartrack-beta.vercel.app/
20•hunzaboy•3h ago•5 comments

Working with AI feels more like leadership than coding

https://allen.bargi.org/notes/working-with-ai-feels-like-leadership/
268•allenb•15h ago•174 comments

Bede Liu, a digital signal processing pioneer, has died

https://spectrum.ieee.org/digital-signal-processing
56•Jimmc414•4h ago•2 comments

Tess's Android Wayland Compositor

https://github.com/wmww/tawc
45•schmorptron•7h ago•5 comments

The complete AI SEO playbook: from zero to 4.6M impressions in 3 months

https://github.com/TraceCohenTech/ai-seo-playbook
3•t_cohen•1h ago•0 comments

Voltair (YC W26) Is Hiring a Test Flight Engineer

https://www.ycombinator.com/companies/voltair/jobs/sSOD2Ox-flight-test-engineer
1•wweissbluth•7h ago

Jaithon 3, a fast programming language with the perfect syntax

https://github.com/abhiramasonny/jaithon
22•AbhiramaVS•4d ago•15 comments

Show HN: Eigendrum - Draw any shape and hear what it sounds like as a drum

https://baselashraf81.github.io/eigendrum/
204•BaselAshraf81•5d ago•96 comments

The Wow signal was a strong narrowband radio signal detected on August 15, 1977

https://en.wikipedia.org/wiki/Wow!_signal
65•firefax•4h ago•16 comments

Pizza Box Project Stack

https://cblgh.org/posts/2026-07-31-pizza-box-project-stack/
32•surprisetalk•4d ago•4 comments

The mathematical beauty of hyperbezier curves

https://linebender.org/blog/hyperbezier/
188•raphlinus•6d ago•37 comments

An image can overflow

https://master.dev/blog/something-nobody-told-you-about-the-image-element-it-can-overflow/
17•ibobev•4d ago•6 comments

Modeling and Verification of Keeta's Consensus [pdf]

https://xescu.re/keeta-consensus.pdf
11•xescure•3h ago•1 comments

Printytron – Describe a part, get a printable STL

https://printytron.com
71•statenjason•11h ago•31 comments

I Remain a Skeptic

https://blog.jsbarretto.com/post/i-remain-a-skeptic
66•lproven•9h ago•42 comments

Humazon

https://hughhowey.com/humazon/
25•herbertl•5h ago•11 comments