HNHacker News
TopNewBestAskShowJobs

hubraumhugo

9,168 karma · joined June 23, 2019

Building web datasets for humans & agents @ kadoa.com

Also made HN Wrapped (hn-wrapped.kadoa.com)

Happy to chat about data, AI, startups, or anything else: adrian@kadoa.com

submissionscomments
hubraumhugo··on Mistral Fine-Tune
Is finetuning still worth it as models advance so fast? Any real-world use cases?

For example, Bloomberg trained a GPT-3.5 class LLM on their financial data last year and soon after GPT-4-8k outperformed it on nearly all finance tasks.

We ended up focusing on having high-quality eval data and an architecture that makes switching to new models easy.

hubraumhugo··on Don't microservice, do module
This parody sums up the issues with mircoservices very well: https://www.youtube.com/watch?v=y8OnoxKotPQ
hubraumhugo··on Working Smarter, Not Harder
I guess having a baby OR starting your own company can have similar effects. The only advantage you have as an early startup is an execution moat, so you have to be relentless in your focus and prioritization.

My unexciting but very effective productivity stack: A website blocker and a single large txt file for notes and todos

hubraumhugo··on GPT-4o
The movie Her has just become reality
hubraumhugo··on Show HN: Hacker Search – A semantic search engine for Hacker News
Great to see cool stuff being built on top of HN data! Many of us rely on this platform as one of the primary sources of information.

> recommending stories to you based on your interests

I built this as a service that monitors and classifies HN stories based on your interests (solved my FOMO): https://www.kadoa.com/hacksnack

hubraumhugo··on Reddit Will License Its Data to Train LLMs, We Made a FF Extension to Replace
Since the blackouts last year and the recent IPO, it feels like astroturfing and spam have increased, while quality contributions have decreased. All usage metrics are up according to Reddit's IPO filings, but it feels like engagement is actually down, or at least lower quality. Many niche subs feel like ghost towns now.

Is this just my subjective impression or do you feel the same?

hubraumhugo··on Programming Is Mostly Thinking (2014)
> how can you experiment with learning on-the-job to create systems where the thinking is optimized?

Best optimization is less interruptions as reasearch shows their devastating effect on programming:

- 10-15 min to resume work after an interruption

- A programmer is likely to get just one uninterrupted 2-hour session in a day

- Worst time to interrupt: during edits, searches & comprehension

I've been wondering if there's a way to track interruptions to showcase this.

[0] http://blog.ninlabs.com/2013/01/programmer-interrupted/

hubraumhugo··on Mixtral 8x22B
It feels absolutely amazing to build an AI startup right now. It's as if your product automatically becomes cheaper, more reliable, and more scalable with each new major model release.

- We first struggled with limited context windows [solved]

- We had issues with consistent JSON ouput [solved]

- We had rate limiting and performance issues for the large 3rd party models [solved]

- Hosting our own OSS models for small and medium complex tasks was a pain [solved]

Obivously every startup still needs to build up defensibility and focus on differentiating with everything “non-AI”.

hubraumhugo··on GPT-4 Turbo with Vision Generally Available
Their naming and versioning choices have definitely created some confusion. Apparently GPT-4 Turbo doesn't just come with the the new vision capabilities, but also with improved non-vision capabilities:

> "I’m hoping we can get evals out shortly to help quantify this. Until then - it’s a new model with various data and training improvements, resulting in better reasoning."[0]

> "- major improvements across the board in our evals (especially math)"[1]

[0] https://twitter.com/owencm/status/1777784000712761430

[1] https://twitter.com/stevenheidel/status/1777789577438318625

hubraumhugo··on Switzerland faces landmark climate-human rights ruling
Switzerland was the most participatory democratic country in the world in 2023[0] and Swiss voters rejected a stricter CO2 law last year[1].

This court ruling seems quite perplexing then, doesn't it?

[0] https://www.v-dem.net/data/the-v-dem-dataset/

[1] https://www.swissinfo.ch/eng/business/switzerland-votes-on-c...

hubraumhugo··on Show HN: I made a discrete logic network card
Super impresssive! I would absolutely love to work on such projects and admire the passion and countless hours that went into understanding and then building the system.

Retirement is not something desirable for me, so maybe that's when I'll spend my time on hardware and software projects like this.

hubraumhugo··on What I think about when I edit (2019)
Any other engineers trying to improve their writing skills? Any additional resources or tips to help me refine my drafting and editing process?
hubraumhugo··on Tool Use (function calling)
> All models can handle correcting choosing a tool from 250+ tools provided the user query contains all necessary parameters for the intended tool with >90% accuracy.

This is pretty exciting news for everybody working with agentic systems. OpenAI has way lower recall.

I'm now migrating from GPT function calls to Claude tools and will report back on the evaluation results.

hubraumhugo··on Show HN: I've built a locally running Perplexity clone
Excellent work! Cool side projects like that will eventually help you get hired by a top startup or may even lead to building your own.

I can only encourage other makers to post their projects on HN and put them out into the world.

hubraumhugo··on When Will the GenAI Bubble Burst?
There is one big difference compared to previous hype cycles:

While crypto or VR tech still hasn't arrived in our daily lives, most of my friends are already using tools like ChatGPT on a regular basis.

Generating images and videos is entertaining, but I believe that the true value of AI lies in automating tedious, repetitive work, both in the digital (e.g. data entry) and physical space (e.g. assembly or logistics work).

It's these "boring" use cases that will bring big efficiency boosts to our economy.

hubraumhugo··on 10 Years of Hacker News "Ask HN: Who Is Hiring" Trends
I'm wondering if there has been an increasingly inflationary use of the term "Senior" in job postings over the past few years. This data somewhat confirms my suspicion.

Anyone else has this impression?

hubraumhugo··on Ask HN: Is anybody getting value from AI Agents? How so?
The term "AI agents" might be a bit overhyped. We're using AI agents for the orchestration of our fully automated web scrapers. But instead of trying to have one large general purpose agent that is hard to control and test, we use many smaller agents that basically just pick the right strategy for a specific sub-task in our workflows. In our case, an agent is a medium-sized LLM prompt that has a) context and b) a set of functions available to call. For example we use it for:

- Navigation: Detect navigation elements and handle actions like pagination or infinite scroll automatically.

- Network Analysis: Identify desired data within network calls.

- Data transformation: Clean and map the data into the desired format. Finetuned small and performant LLMs are great at this task with a high reliability.

The main challenge:

We quickly realized that doing this for a few data sources with low complexity is one thing, doing it for thousands of websites in a reliable, scalable, and cost-efficient way is a whole different beast.

The integration of tightly constrained agents with traditional engineering methods effectively solved this issue for us.

hubraumhugo··on Eternal September
How does HN handle its Eternal September so well? Discussion quality is still great and dang does a great job and introducing newcomers to the HN rules & culture.
hubraumhugo··on Eternal September
Since the blackouts last year and the recent IPO, it feels like astroturfing and spam have increased on Reddit, while quality contributions have decreased. All usage metrics are up according to Reddit's IPO filings, but it feels like engagement is actually down, or at least lower quality. Many niche subs feel like ghost towns now.

Is this just my subjective impression or do you feel the same?

hubraumhugo··on Launch HN: Lumona (YC W24) – Product search based on Reddit and YouTube reviews
We've tried to build this in the past with Looria.com, where we aggreagted and summarized reviews from the most trusted sources, e.g. Reddit: https://www.looria.com/reddit

Couple of challenges:

- Astroturfing is everywhere

- The data sources, especially social media, become more protective with their data

- Monetizing this is super hard. As an aggregator, you're always just the intermediate. The glory times of ads and affiliate marketing are over.

Vetted.ai is working on something similar and they raised $14M in 2022. For all consumers, I really hope one of you will succeed!

hubraumhugo··on Launch HN: Aqua Voice (YC W24) – Voice-driven text editor
Dictation software is huge in the healthcare industry. Every doctor uses it, and a solution like yours could likely make their work much more efficient.

Have you explored this market segment?

hubraumhugo··on Launch HN: DryMerge (YC W24) – Automate Workflows with Plain English
Super cool! I think this could also be a natural progression for ChatGPT and Microsoft Copilot.

With GPT-5, we might see similar capabilities where integrators would simply provide relevant APIs and documentation, while the AI figures out the automation steps and orchestration.

Plugins and custom GPTs were early (failed) attempts in this direction.

hubraumhugo··on Hacker News Rankings. Graphs of HN posts rank on timeline
Nice, lots of cool stuff to do with HN data.

I've recently started analyzing and categorizing all front-page posts with a simple LLM workflow.

The result is an interactive chart that displays the trending topics across different time frames: https://www.kadoa.com/hacksnack/trending

hubraumhugo··on Launch HN: CamelQA (YC W24) – AI that tests mobile apps
As someone who worked in a mobile dev team, I can only applaud your effort! :)

I have a few questions:

As with all new AI-based RPA & Testing frameworks (there are quite many in YC), I'm curious about the costs and performance. Let's say I want to run a few smoke tests (5-10 end-to-end scenarios) on my app across multiple iOS and Android devices with different screen sizes and OS versions before going into production.

What would it cost, and how long would it take to complete the tests?

Do you already have customers running such real-world use cases with it?

hubraumhugo··on 900 Sites, 125M accounts, 1 Vulnerability
Has Supabase learned from this and done a better job?
hubraumhugo··on Grok
When will we reach an upper limit/dimishing returns in terms of number of parameters and mixture of experts?
hubraumhugo··on Show HN: Skyvern – Browser automation using LLMs and computer vision
AI should automate tedious and un-creative work, and data entry tasks definitely fit this description. Rule-based RPA will likely be replaced by fine-tuned AI agents for things like form filling and similar.

Can you share some data on costs and scalability?

At Kadoa, we're working on fully automating unstructured data ETL from websites, PDFs, etc. We quickly realized that doing this for a few data sources with low complexity is one thing, doing it for thousands of sources daily in a reliable, scalable, and cost-efficient way is a whole different beast.

Using LLMs for every data extraction would be way too expensive and very slow. Instead, we use LLMs to generate the scraper and data transformation code and subsequently adapt it to website changes, which is highly efficient.

hubraumhugo··on Launch HN: Glide (YC W24) – AI-assisted technical design docs
Makes sense, it will be intersting to see how the balance between control and autonomy will play out.

From Kapathy on Twitter:

> In my mind, automating software engineering will look similar to automating driving. E.g. in self-driving the progression of increasing autonomy and higher abstraction looks something like:

1. first the human performs all driving actions manually

2. then the AI helps keep the lane

3. then it slows for the car ahead

4. then it also does lane changes and takes forks

5. then it also stops at signs/lights and takes turns

6. eventually you take a feature complete solution and grind on the quality until you achieve full self-driving.

There is a progression of the AI doing more and the human doing less, but still providing oversight. In Software engineering, the progression is shaping up similar:

1. first the human writes the code manually

2. then GitHub Copilot autocompletes a few lines

3. then ChatGPT writes chunks of code

4. then you move to larger and larger code diffs (e.g. Cursor copilot++ style, nice demo here https://youtube.com/watch?v=Smklr44N8QU)

5.... Devin is an impressive demo of what perhaps follows next: coordinating a number of tools that a developer needs to string together to write code: a Terminal, a Browser, a Code editor, etc., and human oversight that moves to increasingly higher level of abstraction.

hubraumhugo··on Launch HN: Glide (YC W24) – AI-assisted technical design docs
> but instead of directly editing your code, it helps you come up with a solid plan before implementing a change.

Is this similar to Devin that just launched today, except that Glide is less autonomous and gives you more control? How do you compare to them? https://twitter.com/cognition_labs/status/176754876313496400...

hubraumhugo··on Claude 3 model family
It feels absolutely amazing to build an AI startup right now:

- We struggled with limited context windows [solved]

- We had issues with consistent JSON output [solved]

- We had rate limiting and performance issues with 3rd party models [solved]

- Hosting OSS models was a pain [solved]

It's like your product becomes automatically cheaper, more reliable, and more scalable with every major LLM advancement. I'm going to test the new Claude models against our evaluation and test data soon.

Obivously you still need to build up defensibility and focus on differentiating with everything “non-AI”.

← PreviousPage 5 of 14Next →