StartupMill (Nocode venture maker)

Building and investing in low code ventures! @JobsinAI @JobsinXR @JobsinFlutter @Acquire_Me Run by @tobyallen007 | Come say 👋

[StartupMill (Nocode venture maker) @StartupMillco]

We’re looking for a creative no cider to join us.

Someone who’s kick ass at branding, websites / webflow and can help define the visual identities of our ventures.

DMs are open or drop their name here...

#jobsinnocode #nocodejobs #nocode

[Jobs in Ai @jobsinai]

🚨 9 BRAND NEW AI JOBS ADDED! 🚨

We’ve just added 9 new AI roles to our site:

Scale AI – Machine Learning Research Engineer, Agents - Enterprise GenAI (San Francisco, CA) https://t.co/ndgYH9Ar9r

See all open AI roles: https://t.co/9iYtKQiKVu #JobsInAI #AIJobs

[Alex Veremeyenko @alex_verem]

RIP prompt engineering ☠️

This new Stanford paper just made it irrelevant with a single technique.

It's called Verbalized Sampling and it proves aligned AI models aren't broken we've just been prompting them wrong this whole time.

Here's the problem: Post-training alignment causes mode collapse. Ask ChatGPT "tell me a joke about coffee" 5 times and you'll get the SAME joke. Every. Single. Time.

Everyone blamed the algorithms. Turns out, it's deeper than that.

The real culprit? 'Typicality bias' in human preference data. Annotators systematically favor familiar, conventional responses. This bias gets baked into reward models, and aligned models collapse to the most "typical" output.

The math is brutal: when you have multiple valid answers (like creative writing), typicality becomes the tie-breaker. The model picks the safest, most stereotypical response every time.

But here's the kicker: the diversity is still there. It's just trapped.

Introducing "Verbalized Sampling."

Instead of asking "Tell me a joke," you ask: "Generate 5 jokes with their probabilities."

That's it. No retraining. No fine-tuning. Just a different prompt.

The results are insane:

- 1.6-2.1× diversity increase on creative writing
- 66.8% recovery of base model diversity
- Zero loss in factual accuracy or safety

Why does this work? Different prompts collapse to different modes.

When you ask for ONE response, you get the mode joke. When you ask for a DISTRIBUTION, you get the actual diverse distribution the model learned during pretraining.

They tested it everywhere:

✓ Creative writing (poems, stories, jokes)
✓ Dialogue simulation
✓ Open-ended QA
✓ Synthetic data generation

And here's the emergent trend: "larger models benefit MORE from this."

GPT-4 gains 2× the diversity improvement compared to GPT-4-mini.

The bigger the model, the more trapped diversity it has.

This flips everything we thought about alignment. Mode collapse isn't permanent damage it's a prompting problem.

The diversity was never lost. We just forgot how to access it.

100% training-free. Works on ANY aligned model. Available now.

Read the paper: arxiv. org/abs/2510.01171

The AI diversity bottleneck just got solved with 8 words.

[Ilir Aliu @IlirAliu_]

Every robot you see is a data firehose generating terabytes of chaos.

This hidden crisis is the #1 reason robots fail, and it's costing the industry billions.

You see hardware, but not the data swamp drowning engineers.

In 2025, a quiet revolution is fixing it. Here’s how. 🧵

[StartupMill (Nocode venture maker) @StartupMillco]

I have a deep hatred of debugging Sendgrid dynamic emails.

I’m parsing json and only some of the JSON (thanks to @minimumstudio’s great json builder) is getting picked up.

Try the output data in Sendgrid’s preview and it all works. Try the sandbox tester and comes back all ok.

[StartupMill (Nocode venture maker) @StartupMillco]

@bentossell @levelsio To be fair, he’s got a big audience to launch with.

[StartupMill (Nocode venture maker) @StartupMillco]

@yelkhayami @minimumstudio @linear Oh man… linear..

Look at Clickup.

[StartupMill (Nocode venture maker) @StartupMillco]

@PhilipLakin Spent a long time with airtable as a PM tool and would not use it again but really depends here on the jobs to be done which would need a deeper dive.

Probably recommend to go full bespoke with bubble as long as the processes are well defined.

[StartupMill (Nocode venture maker) @StartupMillco]

@yelkhayami @linear Talk to https://t.co/vzCUz3aOIS

[StartupMill (Nocode venture maker) @StartupMillco]

It’s been a while..

[StartupMill (Nocode venture maker) @StartupMillco]

@yelkhayami @bubble Yeah still figuring out how tho

[StartupMill (Nocode venture maker) @StartupMillco]

@eyishazyer @tobyallen007

[StartupMill (Nocode venture maker) @StartupMillco]

@bentossell Highly likely you triggered an ai / fraud / political advertising protection system and you need to get unflagged

[StartupMill (Nocode venture maker) @StartupMillco]

Yup we feel you! Love this!

Was looking at timings once me side and from start to deploy, takes two minutes to launch a full feature.

[Kieran ⚡️ Building apps with AI + nocode 🦾 @nocodelife]

Q: How fast is no-code?
A: It took me 5 days to make this SaaS product from scratch.

Features:
- User authentication
- Premium plans
- Stripe subscriptions
- Page customisation
- User custom domains
- Search + filter page
- Account mgt

I could have done it faster but...

[StartupMill (Nocode venture maker) @StartupMillco]

Any #nocode or #lowcode folks hitting up #websummit2022 then reach out to @tobyallen007 as he’ll be there!

[StartupMill (Nocode venture maker) @StartupMillco]

@DruRly We have been with @JobsinNextTech for xr, quantum, sextech, gaming, ai, flutter, crypto

[StartupMill (Nocode venture maker) @StartupMillco]

@yelkhayami Try it with https://t.co/QekahwbsKw and it’s super powerful!

[StartupMill (Nocode venture maker) @StartupMillco]

@ClausISaidSo @stripe @bubble Happy to help

[StartupMill (Nocode venture maker) @StartupMillco]

@nocodechris @bubble We’ve found the same

[Toby Allen 🦄 @tobyallen007]

Looking for #bubble developers for two projects

#jobsinlowcode

[StartupMill (Nocode venture maker) @StartupMillco]

We are looking for a Sr Coms and Marketing person to help us scale our messaging!
#startupjobs