frontpage.
newsnewestaskshowjobs

Open Source @Github

fp.

Open in hackernews

Show HN: Qwen3-4B with the AI writing tics trained out

https://100toolkit.com/blog/deslop-model/
57•erikkannike•1h ago

Comments

anilgulecha•1h ago
What's with the very distracting cursors of others?
mullenba•59m ago
Exactly. I was interested in the article but I don't have time for that nonsense.
markgavalda•58m ago
Agreed. In the upper right corner there's a "hide" button that I immediately clicked. Then read (ok, skim read) the article.
HumblyTossed•41m ago
Agreed. Just because you _can_ do a thing, doesn't mean you should.
awestroke•58m ago
Nobody knows what it means, but it's provocative. It gets the people going
Almondsetat•57m ago
going... away?
maven29•55m ago
The ancients brought us "reading mode" in browsers as an escape hatch for exactly this kind of cruelty
jmusall•49m ago
There's a "hide" button in the top right corner
2bird3•40m ago
All it is missing now is a chat and a friend/follow feature.....
RonanSoleste•40m ago
I don't understand why this is a feature and what it brings other than shout "you are being watched by random others". Exited the site before ending the first block.
ac2u•47m ago
Is it still capable of writing an essay about AI writing tics or has the very means of producing said tics, even when the model is asked to do so to demonstrate, now abitterated?
swiftcoder•40m ago
Impressed how little hardware is needed to train this.
hashmap•35m ago
I saw those moving dots and snap-closed the page.
chmod775•32m ago
Models have proclivities, and many people use the same model.

Imagine all books on earth were written by just 5 people - and worse, they could not remember what they had written yesterday. You would quickly get tired of reading the same flourishes and patterns over and over.

Whatever style models currently in use have, that's what people are going to be tired of, and it's what they're going to recognize as "AI".

sejje•23m ago
I predict by the end of the year they'll have a style/voice selector for responses.

You can already do that in your agents.md file, of course, or ask for a re-write and dodge the sameness.

wongarsu•12m ago
The most difficult thing in doing it yourself is knowing what to ask for. I can ask my model to talk like a caveman or like a Baltic pirate. But asking for something more nuanced, in the range of regular speech, requires the correct lingo

I'm sure it's trivial for somebody who has studied communication or contemporary linguistics, just as it's trivial for me to ask an LLM for tenant-isolation using row-level security in an app

Royce-CMR•22m ago
Applause earned. Neat results especially out of such a small model.

Separately we have overlap between AI slop styles and right for the situation writing styles which this highlights.

Using the new CRM example, the clean version looks like a millennial student who rambled on a page, and the AI slop version looks like a draft I could send for PR to corporate managers. I buy the former looks more real, but it also looks sloppy and doesn’t work for most of my use cases.

When I was in school, the structured essay style of the “slop” examples was encouraged and I see it continue to resonate in corporate. The cleaned up versions here def have use cases - reviews for example.

Main point - different styles for different situations and our AI slop versions vs humanized versions both overlap different areas of the “ideal” we continue to look for.

trilogic•20m ago
The q4 loops after 16k tokens generated. Great idea though, hope to see more models like it. Thanks. https://huggingface.co/erikkannike/qwen3-4b-deslop-v4/discus...
charv•20m ago
Side note but the multi-user cursors was a super cool feature, not sure it was relevant to this article but would love to see it on more interactive pieces (a la The Pudding, etc.)
JustSkyfall•16m ago
It's definitely a slight improvement, but it is still rather obvious that the output is AI-written :(
mrweasel•13m ago
The tell tell sign for me is that it's still incredibly hard to read. AI tend to write in a style that I'm unable to easily read (humans does that as well sometimes). I constantly have to go back and re-read the previous sentence, because the text simply doesn't register.

A next step I'd suggest is asking the LLM to use shorter words when available.

BoredomIsFun•10m ago
Did you check this? https://huggingface.co/sam-paech/gemma-3-27b-it-antislop
BoredomIsFun•15m ago
This is not the first deslop. The true pioneer is this dude: https://huggingface.co/sam-paech/gemma-3-27b-it-antislop. This model is a perfectly usable, full size not-toyish model.
f311a•6m ago
Why would you support this sort of projects? LLM text should be spottable for as long as possible, otherwise human communication and expression on the internet will be killed.
10111001•5m ago
Now can I have a model with the AI writing tics trained to the max?

I want my boss to know I'm a Good Boy using AI.

BSOhealth•3m ago
for https://slopwash.com I’ve started approaching this more as an eval problem than something to go post-training for. I’m excited to be trying these new Jev-type models in more of a harnessed approach. maybe some heavy LLM to measure preservation of intent

Cop Fired for Letting Her ICE Agent Husband Use Her Flock Login

https://gizmodo.com/cop-fired-for-letting-her-ice-agent-husband-use-her-flock-login-2000827428
1•pavel_lishin•2m ago•1 comments

Fishfolk/bones: An easy-to-use game engine for making real games

https://github.com/fishfolk/bones/
1•janandonly•2m ago•0 comments

Evaluating an AI sales agent on 100 sites/ 15M visitors (+10% rev per visitor)

https://getjavo.com/research/sales-evaluation/
1•benataiso•4m ago•0 comments

Eval() in a straitjacket: sandboxing JavaScript under CSP

https://stringutilitybelt.com/blog/eval-in-a-straitjacket/
1•Murraylr•4m ago•0 comments

What China Gets About AI That America Doesn't [video]

https://www.youtube.com/watch?v=iorTxXfBlMQ
1•hunglee2•4m ago•0 comments

Claude Audits Quantum Gravity

https://github.com/eokahya/claude-audits-quantum-gravity
1•iphonecorridor•6m ago•0 comments

Nvidia-Backed Startup DoubleAI Launches Model to Pare Cost of Compute

https://www.wsj.com/cio-journal/nvidia-backed-startup-doubleai-launches-model-to-pare-cost-of-com...
1•ronfriedhaber•6m ago•0 comments

Show HN: A5 v1.0

https://a5geo.org/blog/2026/10/14/a5-v1
1•pheelicks•8m ago•0 comments

Japan Turned Department Stores into Dreamworlds

https://thereader.mitpress.mit.edu/how-japan-turned-department-stores-into-dreamworlds/
1•sohkamyung•9m ago•0 comments

Hanami, Why?: Model Citizens

https://aaronmallen.me/writing/hanami-why-model-citizens
1•aaronmallen•9m ago•0 comments

Beyond These Stars

https://store.steampowered.com/app/2295060/Beyond_These_Stars/
2•doener•12m ago•0 comments

Snowflake data metric functions can now skip transient tables [video]

https://www.youtube.com/shorts/Ar-RWgNReII
1•yanivleven•13m ago•0 comments

We tried to levitate The Pentagon: Ed Sanders trickster god of US counterculture

https://www.theguardian.com/music/2026/oct/19/ed-sanders-interview-the-fugs-film-counterculture
1•rdmuser•14m ago•0 comments

Learning to work faster than doubt

https://thecreativeindependent.com/people/musician-molly-nilsson-on-learning-to-work-faster-than-...
1•herbertl•14m ago•0 comments

Local OpenRouter proxy that picks a provider by cost and speed

https://github.com/mikael-titinovskii/orr
1•kaunakid•14m ago•0 comments

All vulnerabilities are now race conditions

https://blog.himanshuanand.com/2026/09/all-vulnerabilities-are-now-race-conditions/
1•unknownhad•15m ago•0 comments

The significance of General Tso's chicken (and pizza)

https://loveofallwisdom.com/blog/2026/10/the-significance-of-general-tsos-chicken-and-pizza/
1•herbertl•15m ago•0 comments

Show HN: Free text classification for your n8n workflows

https://jeffyclassify.com/blog/n8n-jeffy/
1•nico•15m ago•1 comments

How Africa's 'wall' to hold back the desert unravelled

https://www.ft.com/content/13cbc4f0-6c8f-4b95-bbe8-7eb03a547727
1•bookofjoe•15m ago•2 comments

In Praise of Ignorance

https://jamesmiller411396.substack.com/p/in-praise-of-ignorance
1•herbertl•16m ago•0 comments

Show HN: Footloom – Self-Hosted Google Analytics Alternative for WordPress

https://wordpress.org/plugins/footloom/
1•hamzamairaj•17m ago•0 comments

Archive 2012: The Whispering Earring

https://web.archive.org/web/20121008025245/http://squid314.livejournal.com/332946.html
1•morgangiraud•17m ago•0 comments

A Rose.com by Any Other Name (2001)[pdf]

https://home.business.utah.edu/finmc/FinalJFversion2371-2388.pdf
1•mday404•17m ago•1 comments

Halifax Leaked My Financial Information (Again)

https://danq.me/2026/10/09/halifax-dun-goofed-again/
2•surprisetalk•18m ago•0 comments

ChatGPT Bulk Delete

https://chatgptbulkdelete.net/
1•vinghunter•21m ago•0 comments

Ask HN: Testing Text Posts

1•njblue•21m ago•0 comments

Show HN: KiviDB, a faster Redis-compatible memory database and vector cache

https://github.com/kividbio/kividb
2•murtazayusuf•24m ago•1 comments

Why Is Speculative Decoding Fast?

https://jbarrow.ai/2026-10-08-why-is-speculative-decoding-fast/
1•gmays•24m ago•0 comments

Canonical's Multipass 1.17 Adds Direct Support for Non-Ubuntu Linux Distros

https://www.phoronix.com/news/Multipass-1.17
1•mikece•25m ago•0 comments

Dr Garry Hunt talks to us about the astonishing Voyager probes

https://www.theregister.com/offbeat/2026/10/19/dr-garry-hunt-talks-to-us-about-the-astonishing-vo...
1•sohkamyung•25m ago•0 comments