What are you working on? Any new ideas that you&#x27;re thinking about?

Ask HN: What are you working on? (June 2026)

TLDR: I had 2,207 GoPro videos, and I need to rewatch them to find interesting moments from my cycling journey. I built a project to index them locally on my M1 Max using open-source ML models, search for those moments, and send the best clips straight to my DaVinci Resolve timeline. I indexed 628 videos (668.68 GB, 15h 13m 18s of footage duration), more details in the metrics table in the last section of this article.Full article: https:&#x2F;&#x2F;iliashaddad.com&#x2F;blog&#x2F;i-indexed-669-gb-of-my-gopro-videos-using-my-m1-max-computer

I indexed 669 GB of my GoPro videos using my M1 Max computer and local ML models

FTX held a diluted 7.84% stake in Anthropic, according to Reuters.Anthropic’s latest reported valuation is around $965B.That implies the former FTX stake would be worth about $75B before further dilution.FTX’s customer shortfall was roughly $8B to $9B.The estate sold the Anthropic stake during bankruptcy to repay creditors.Sources:
https:&#x2F;&#x2F;www.reuters.com&#x2F;technology&#x2F;crypto-exchange-ftx-sell-shares-ai-startup-anthropic-2024-02-22&#x2F;
https:&#x2F;&#x2F;www.reuters.com&#x2F;technology&#x2F;openai-files-us-ipo-after-anthropic-ai-giants-head-public-markets-2026-06-08&#x2F;

FTX's former Anthropic stake would be worth about $75B at today's valuation

Claude and ChatGPT have thinking efforts where you can tune the amount of thinking allowed.Like low, medium, high, xhigh and so on.But are they different models underneath? Or same model with different parameter?The reason I ask is because, if I change the effort param mid conversation in Claude code, I get a warning suggesting I’m breaking the cache.I don’t think this happens in Codex because when I change the effort, the responses are still quick.

Ask HN: How are thinking efforts implemented?

This is a stupid example to illustrate what I mean. Say you have this code:<pre><code> def create_background(width: int, height: int) -&gt; Image:
 ...
</code></pre>
You tell the agent to use default values for create_background, the same as in create_screen. It changes the code to:<pre><code> # Now create_background params have default values, the same as create_screen in screen.py
 def create_background(width: int = DEFAULT_WIDTH, height: int = DEFAULT_HEIGHT) -&gt; Image:
 ...
</code></pre>
The unnecessary comment is a staple of vibed code, but the tone also annoys me because it leaves behind the prompt. It words comments based on what it was asked to do, not in a timeless manner.I keep telling people in code reviews to remove unnecessary comments, and I feel I lack the vocabulary to express why this is bad.

Is there a name for the type of comments agents add where they leak the prompt?

Hey gang, you may remember me from such books as _The Lean Startup_ and _The Startup Way_.It&#x27;s been fifteen years since I wrote The Lean Startup, and in that time I&#x27;ve seen some things. In both big companies and tiny startups, NGOs and governments, in almost every industry you can name.I&#x27;ve helped a lot of people create a lot of amazing companies, but I&#x27;ve also seen so many ways this can go wrong. There&#x27;s a darkness in our industry that we often don&#x27;t talk about.I kept watching good companies drift away from the missions they were founded on. Not because anyone woke up one day and decided to be evil, but because the structure they were built on slowly pulled them there. I call that pull &quot;financial gravity.&quot;We&#x27;ve all experienced watching a company we love or admire be warped and broken beyond recognition; until it&#x27;s a husk of its former self, or worse. I wanted to understand why. And I wanted to know what all of us can do to stop that from happening.My new book _Incorruptible_ is my attempt to explain the invisible forces that shape organizations, and how a handful of companies (like Costco, Patagonia, and Novo Nordisk) have successfully been structured to resist gravity and thrive for decades -- or even centuries.Along the way, I founded the Long-Term Stock Exchange, co-founded an AI R&amp;D lab called Answer.AI with Jeremy Howard, and helped a number of notable companies with their governance (yes, including Anthropic).I won&#x27;t pretend I have this all figured out, but I&#x27;ve probably spent more time than is healthy on the &quot;why do good companies go bad&quot; question. Ask me anything!

I'm Eric Ries, author of "The Lean Startup" and new book "Incorruptible" – AMA

After ten years I’m working on a redesign of my personal tech blog. I want to look at some other blogs with good modern design. Of course searching for ‘well designed blogs’ or ‘webdesign 2026’ brings up endless marketing listicles. Even after wading through the crap, the advice is focused on advertising and product centric sites, not prose heavy sites with good typography.So my question to you is: What are some of your favorite text heavy blogs that are a joy to read?I don’t mean the content (though that helps too), but rather sites where the actual experience of reading is pleasant. Think: good fonts, sticky headers (or not), well formatted code snippets, and responsive images. Slide in navbars vs inline table of contents.Thanks,
 - j

Ask HN: Favorite text heavy blogs that are a joy to read?

It feels like every event&#x2F;venue is selling tickets exclusively through Ticketmaster.  Every other ticketing platform seems to only hold resale tickets in their inventory which just transfers the tickets to your Ticketmaster account when bought. With all the hate Ticketmaster has gotten and all the other ticketing platforms out there, I&#x27;m surprised Ticketmaster still has a hold of pretty much the entire market. How are they doing this? Why haven&#x27;t the other platforms been able to compete?

Ask HN: Why hasn't there been a real competitor to Ticketmaster yet?

&gt; For Fable 5, Mythos 5, and future models on Bedrock with similar or higher capability levels, Anthropic will require 30-day retention for all traffic on Mythos-class models. Retaining data for a limited period allows Anthropic to detect patterns of misuse that are not visible from a single exchange. Once you opt into data retention, your data will leave AWS’s data and security boundary.From the announcement here: <a href="https:&#x2F;&#x2F;aws.amazon.com&#x2F;blogs&#x2F;aws&#x2F;anthropic-claude-fable-5-on-aws-mythos-class-capabilities-with-built-in-safeguards-now-available&#x2F;" rel="nofollow">https:&#x2F;&#x2F;aws.amazon.com&#x2F;blogs&#x2F;aws&#x2F;anthropic-claude-fable-5-on...</a>&gt; After 30 days, the data is deleted automatically, except in the rare cases where it&#x27;s part of a safety investigation or we&#x27;re legally required to keep it.From: <a href="https:&#x2F;&#x2F;support.claude.com&#x2F;en&#x2F;articles&#x2F;15425996-data-retention-practices-for-mythos-class-models" rel="nofollow">https:&#x2F;&#x2F;support.claude.com&#x2F;en&#x2F;articles&#x2F;15425996-data-retenti...</a>

AWS Bedrock to require sharing data with Anthropic for Mythos and future models

Not a hiring post. I have a 9-5 but also a lot of ideas I want to build. The core is just to show up and make small progress — but I find I&#x27;m missing accountability and partnership, so I thought it&#x27;d be nice to partner with someone. Mostly backend&#x2F;infra stuff. If you&#x27;re interested, email in my profile.

Ask HN: Want to build something open source on nights and weekends together?

Notes on DeepSeek

We should start measuring knowledge debt like the way we do for tech debt

In these modern harrowing times, more and more posts are proving to be AI slop, and many people are averse to that. Do you think it would it be useful to have a way to indicate that you believe a post is slop separate from the existing flag button? Maybe it&#x27;s just me but I feel like having a dedicated [AI slop] flag on a post would be useful because I feel differently about posts that have been flagged for other reasons than that. (or something friendlier if value is seen in the general idea but not in that language)

Ask HN: Would it be useful to have a slop button in addition to flag?

This submission is a tale about how I launched an unlimited LLM provider to about 60 hyped people on the waitlist, then immediately served them a fully dysfunctional death-loop model, and how most people, very reasonably, disappeared, but thanks to a few extremely nice people stuck around anyway, we kept the project alive and its still pretty chaotic but gaining traction.To back up a little bit-- I believe that the whole point of AI agents is that they should keep working. They should read files, retry, search, code, summarize, run tools, and loop until the job is done. When your employer is paying for it, who cares about cost, but when it comes to my personal money&#x2F;hobbies, if every loop feels like a tiny financial event, you start babysitting the agent instead of using it, and its not fun.On the other hand, metered pricing makes me worry about using too much. Usage subscriptions make me feel like I need to use every last magical % or I&#x27;m are &quot;wasting it&quot;. If only an unlimited provider existed....Then I joined the AMD developer program - I got some credits to spin up my own MI300x and started tinkering with vllm&#x2F;sglang inference serving on AMD.After learning about AMD MI300x , i did some napkin math:Renting MI300x at 2.00 an hour = ~$1500 a month . It can probably support about 150 users using a small MOE model, like qwen-35b-3a , maybe more.1500 &#x2F; 150= $10.00 per month, and we all get to play with agents for a small price.You can oversubscribe a bit, so i landed on $6 per month, per user, for 2x generation slots, 128k context, no token limits, no rate limits.I built the site, router, made a waitlist, and then over-optimized the MI300x to the point where vllm bench had like 3k+ output and 40k+ throughput.... But i didn&#x27;t test the final config&#x2F;serve commands... And that&#x27;s where i did a disaster launch. You couldn&#x27;t prompt the thing without it looping or bugging out, it was cursed. And that&#x27;s where we lost alot of people.Luckily, my buddy had a few 3090s, so he threw me a life boat and began hosting qwen for us on 2x 3090s and we finally had an operational model that wasn&#x27;t costing $2.00 an hour for our whopping 3 users.We started gaining a more users, so we moved up to 4x 3090s. Which we have plenty of room for more users, but even so, since then:we&#x27;ve configured vllm wrong like 15 times
a GPU died
we lost power
I made a bunch of one-click starts for openclaw,hermes,pi-mono and none of them really work right and that probably drives people away. Those are still on our site right now....but people that know what they are doing seem to really be liking the price point. All in all we have like 98% up time. Its been about a month. We&#x27;ve both learned a ton, even already having backgrounds in SWE&#x2F;SE&#x2F;AI , being on the hook for a couple paying users forced us to really focus on delivering them a good product. And now i think we might be close to paying the power&#x2F;hosting bill so we&#x27;re not operating at a loss (if u include 3090 capex were still at aloss).Our break-even point is moving to the cloud to max out a MI300x, which is now tuned and ready to go once we get the users.And im finding in some areas, subscribing to our service is cheaper than running the model (but as someone who loves local models, i totally get it).Since then, I&#x27;ve been working on a desktop agent that actually works with small models like qwen -- thats going to replace the broken 1 click starts. It&#x27;s barebones, but its something out of the box that just works. I made it open source, you can see what im talking about here: https:&#x2F;&#x2F;github.com&#x2F;yolo-auto-org&#x2F;yolo-auto-desktop , we&#x27;re at yolo-auto.com and we have an abysmal free tier to prove it works!Anyway, hope you got a laugh or found it interesting! Drop a question if you have any.

Story of How Im Running an Unlimited $6/Month AI Provider on 4x RTX 3090s

The large companies I’ve worked at (including FAANG) seemed to thrive on kudos via performative actions. Like the majority of the team is doing useless stuff that management thinks is impressive while the couple all stars get the team closer to the goal.Meanwhile, a lot of managers calendars are purely just 1:1s with devs on the team which clearly has very little value add to the team.Anyone else notice this? Not sure if there’s a word for it, but it’s somewhat demoralising working with a bunch of corporate office workers cosplaying as engineers

Ask HN: Are most corporate SWE jobs performative?

Ask HN: Why is the best way to find a job as a Software Engineer in 2026?

Before agentic coding, I always prided myself on how long I could work in a flow state. I was really good at working deeply.Now, with slow agents like Claude, I find myself no longer working deeply.What are you all doing to stay focused?

Ask HN: How do you get into a flow state when using AI to code?

Ask HN: What are tools you have made for yourself since the advent of AI?

Hello, fellow overstimulated kidsI don&#x27;t know if it&#x27;s something of my generation, my worsening ADHD or just laziness, but whenever I sit down and start studying something, magically, I find myself researching the most random shit ever.Need to do at least 1 hour of Game Theory? Best I can do is 3.5 hours of WWII fun facts.Anyways, as the lazy person I am, instead of finding the willpower to focus and get shit done, I made a customizable forward proxy. You can define your custom rules in nodes, arrange them visually in policy flows and bundle them into modes. You can find it here: https:&#x2F;&#x2F;github.com&#x2F;Vaccarini-Lorenzo&#x2F;ProductivityProxyIt&#x27;s a project I built for personal use but, with work and shit, I don&#x27;t have as much free time as I used to, so I hope someone smarter and with more time than me might just take it and make it good. Or maybe it&#x27;s just useful to someone as it is.A few notes:- It&#x27;s a project thought for &quot;power-users&quot;, you might need to write some (python) code in the nodes to describe their behaviour, but hey, just let a LLM do it for you: The project is well documented and I plan on bundling an AGENTS.md and expose agent-friendly APIs so that you can define your local flows in plain natural language and let your favourite LLM write the policies.- It is built on top of mitmproxy, you will need it as dependency (as well as their certificates)P.S.
It&#x27;s a Tauri app. Right now the automatic proxy setup is macOS-only (I&#x27;ve tested on Apple Silicon). On other platforms you&#x27;d have to point your browser at the proxy manually, and cross-platform support is still a TODO.

I procrastinate by building tools to stop me from procrastinating: A sad story

Almost two years ago there was a thread on this (https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=40872102).  I&#x27;m curious now that more time has passed what people think?

Ask HN: Are you still using a Vision Pro?

In the last one or two months, starting from DeepSeek V4 Pro, there are quite many low-price Chinese models coming out. Their performance looks more or less similar to me: Mimo V2.5 Pro, MiniMax M3, and the just released GLM 5.2, etc.Which model are you using now? Why? What are the good and bad parts?

Ask HN: Which cheap Chinese LLM are you using?

I and a small cohort of friends truly believe that the &quot;$1T&quot; floats are going to burn capital. We think this is a disaster of landgrab spend, and that the cost of a token cannot be recovered by future commercial use.If this theory is correct, then following on from the housing debt crisis, this is a hollow shell and so somebody in finance should be shorting the floats, and betting on a crash to clean up.Polymarket betting aside, is anyone able to confirm that there are funds looking at a short play in the AI space, or is everyone on-board with things and nobody is taking the other side of the bet?I am not in fintech or investing, I am not seeking alpha to do this, I don&#x27;t invest directly. I am however interested in validating my theory that somebody smart out there sees a huge upside in the explosion when this attempt to buy market share in Unicorn juice explodes.NVDA as an example has under 2% of its trade in shorts. I can&#x27;t tell if thats big or small TBH. Feels small.

Ask HN: Is anyone shorting the overspend in AI yet?

I don’t know if it’s just me, but yesterday’s US decision to ban access to the Fable model feels like an epochal shift in the “AI race,” something on the scale of the Trinity test.It is hard to count how many boxes were ticked yesterday:A government shutting down an AI model it doesn’t like? A government revoking tens of billions of dollars in revenue from a barely profitable $1T startup whose entire trajectory, and in essence its survival, depends on the commercial success of that model? Access to an AI model being governed by citizenship, likely verified through rigorous ID checks?Imagine that the model is made available again next week. Can we trust it, or trust the changes imposed by Anthropic to comply with the US? Likely not. Can we trust that the US government will not have access to the non-restricted model for its own cybersecurity operations? Likely not.There is little doubt that China will start to follow suit. We already see Chinese companies slowly scaling back their openness, and there are rumors that this trend will continue. Now we can expect Chinese companies to start releasing redacted models or limiting access based on nationality or location.It feels like the US government opened Pandora’s box yesterday and pushed us into the territory of a weaponized AI race, with more restrictions, control, and regulation of access. Even if, in the end, this is all another TACO story or a “marketing campaign” from Anthropic, the damage is very unlikely to be undone.

Ask HN: Did we witness the "Trinity moment" for AI?

This made a huge difference for me. Try it out!

Tell HN: iOS devs, get back lots of disk space: xcrun simctl delete unavailable

I was just wondering if some variation of this idea has been tried before?

What if we legally required politicians to work regular jobs 2 months a year?

I use Claude daily and generally not a detractor to reasonable use of AI where it makes sense.However, when executive pushes AI onto an org it can feel like being told what compiler&#x2F;IDE to use by people who have no idea about engineering, maybe there is something I&#x27;m not seeing tho, so my question...What is the situation in your org and how is AI adaption driven at your company? I wonder what the split between companies adopting engineering driven approach and pray and spray approach to AI adoption is...

Ask HN: What is the AI adoption approach at your org?

the title is the question. that&#x27;s all

Did anyone went to YC directly from Sri Lanka?

I&#x27;ve tried many different models and without doubt the code coming out of them differs a lot when it comes to &quot;quality&quot;. Some of that is subjective for sure, but there are objective sides to &quot;good&quot; code.I wish this was a metric for the AI benchmarks so I could choose a model based on this, because honestly it&#x27;s one of the things I care most about.Problem: How can you measure such things, whats the metrcis?...maybe there just isn&#x27;t a way to do it, since that metric isn&#x27;t in the charts..

Ask HN: Is there a metric for AI code quality?

I have a very high-quality SaaS product that&#x27;s ready to launch. Everything is complete: the product, testing, LLC, bank account, Stripe, tech stack, etc. But I built it as a programmer trying to create the coolest thing possible for a product I&#x27;ve always wanted to make.
My issue is that using Claude as a lawyer is scaring me. What I consider moral and what seems to be legal are two different things.
For context, the product uses an AI pipeline to adapt user-provided content into dramatized formats. Think of it as ebook-to-graphic-audio adaptation (though there are other pipelines). The quality is insanely good, but having users bring copyrighted content they own to be adapted is legally unclear. ElevenLabs doesn&#x27;t add sound effects, so they&#x27;re fine, but this involves transcription plus additional steps. Claude&#x27;s legal responses are also scaring me. Even with aggressive content moderation to reject CSAM-related materials, report to authorities, block users, and include it in the ToS, there&#x27;s a risk that some things could slip through and I&#x27;d get sued for massive damages.
I also need clarification on whether I&#x27;m allowed to cache and reuse generated assets. For example, if a sentence like &quot;Pizza is good&quot; is generated once, can I reuse it a second time at a lower cost instead of regenerating it?
I&#x27;m just a single programmer without infinite capital to license content. Public domain and CC work won&#x27;t drive the user interest I need.
I need real legal advice, but I can&#x27;t justify spending $5k on a lawyer just to hear this is a bad idea. What can I do?

Ask HN: I Need Help for a Product

Ask HN: What internal tool did you build that became a product?

Yesterday, I got a call from 646-257-4500.American western male voice. Very polite. They actually called me 3 times. The first two, I just hung up.They were claiming they received a request from the google support portal for a change of phone number on my account and wanted me to verify my account.They sent me an email which looks very much like it came from Google… even in the headers! I don&#x27;t see anything intrinsically wrong in it.Subject: Re: You are now on the phone with a verified Google Agent, your Case ID is: XXXXX. Please ask your Agent to confirm this over the phone.<pre><code> ARC-Authentication-Results: i=1; mx.google.com;
 dkim=pass header.i=@google.com header.s=20251104 header.b=XXXX;
 spf=pass (google.com: domain of XXXX.XXXX.XXX@cases-outbound-prod.bounces.google.com designates 209.85.220.75 as permitted sender) smtp.mailfrom=XXX.XXX.XXX@cases-outbound-prod.bounces.google.com;
 dmarc=pass (p=REJECT sp=REJECT dis=NONE) header.from=google.com;
 dara=pass header.i=@gmail.com
 Received: from mail-sor-f75.google.com (mail-sor-f75.google.com. [209.85.220.75])
 by mx.google.com with SMTPS id XXXX-XXXX.10.2026.06.11.14.42.06
 for &lt;XXXX@gmail.com&gt;
 (Google Transport Security);
 Thu, 11 Jun 2026 14:42:06 -0700 (PDT)
</code></pre>
They hung up immediately when they realized that I wasn’t going to read them that code.Searches for the number confirm I&#x27;m not the only one.I guess my question is how they could send that email!?Why isn&#x27;t google filtering this out through their system?

Ask HN: Phishing from 646-257-4500

My first submission to the Apple App Store has been rejected due to &quot;Spam&quot;.According to the guidelines:
&quot;
- Submitting an app with the same source code or assets as other apps already submitted to the App Store
- Creating and submitting multiple similar apps using a repackaged app template 
- Purchasing an app template with problematic code from a third party 
- Submitting several similar apps across multiple accounts
&quot;I built my app 100% from scratch - including all code, media assets, design, copywriting.
Everything was built by me.
There is only one dependency which is the AI model weights which runs locally on device... can&#x27;t expect those to be unique too!I have only one submission in this one and only account.I assume they automatically flagged the app&#x27;s AI model weights - but that should not be flagged!How can I handle this so I can eventually get my app approved?Because they didn&#x27;t even reference to any other similar apps. Does it even matter? The store is flooded with clones of TODO apps, calorie trackers, etc.
I believe my app is actually unique.

Ask HN: How to deal with App Store rejections?

I&#x27;ve noticed an effect with openai where my codex agents seem to perform worse in the week(s) leading up to a new release. I&#x27;m wondering if the vendors tweak effort params at all to free up hardware to host the new version. It&#x27;s a double win as the new model will look night and day better to their regular users when the new model is released presumably with effort back at normal levels.Is this a known phenomenon? Are any folks trying to measure any of this objectively?

Ask HN: Agents get dumber before release of new model version?

Ask HN: What is your blog and why should I read it?

I watched the movie Backrooms last weekend and found it enjoyable. The biggest thing that jumped out to me was how similar it resembled Legend of Zelda: Ocarina of Time temple&#x2F;dungeon layouts. Did anyone else have similar thoughts?There was an early scene where the Backrooms character had to kneel down and crawl into a small knee-high opening leading into another room.There were multiple scenes where a &quot;grinning sun&quot; wall figure is shown. There were some scenes where the rooms expand &quot;into the wall&quot; in a rotated way. And the perfect cubic cutouts of the floor and ceiling that occasionally lead to another room.These were all the major things from what I can recall and almost feel like direct &quot;ripoffs&quot; from Ocarina of Time. I felt they were well done and have no intent of trashing the movie. While watching it just felt the director was probably a big Zelda fan.Curious if anyone has thoughts here or if these are just coincidental stylistic choices.

Ask HN: Influence of Legend of Zelda in Backrooms Movie?

(I don&#x27;t work at Google. I just hope some Googler reads our answers and submits feature requests based on them :-)I often write documents in Google Docs on my laptop. Every time I open a new document, I change the font from Arial to Lexend and set the size from 11 to 14 points. I wish Google Docs would just remember my favorite font.Sometimes I insert special characters into my documents. The Insert &gt; Characters &gt; Special characters window is handy. I like that I can actually draw the symbol I&#x27;m looking for. However, I have an idea that would make inserting special characters even easier: adding a few thumbnails of recently used special characters to the toolbar.I also miss having a button on the toolbar to toggle dark mode with a single click. To be clear: I am talking about Google Docs for desktop.What features do you feel are missing in Google Docs?

Ask HN: What features do you miss in Google Docs for desktop?

Ask HN: What would you do with a trillion dollars

Developers working with LLMs in a professional environment, is Spec Driven Development really the way forward with LLM coding agents?Over the past few months I&#x27;ve really tried to lean into SDD, after trying SpecKit and GSD, I finally settled on OpenSpec as my preferred spec generator.However I&#x27;m not quite ready to give up on understanding the actual code that the LLM is producing.To help with maintaining my comprehension of the generated code, I&#x27;ve tried a few different methods, such as generating smaller specs for more granular features, and pausing after each major task block to review the code. Approaching this via &quot;&#x2F;ospx-apply Run until task 3.5 is complete and then pause for review&quot; gives me the ability to vet the code being produced and make corrections if needed. But honestly, the major issue that I have is that coding style and good code hygiene is often lacking, even with appropriate AGENTS.md and Skills loaded (the usual issues of multiple helper methods, inconsistent naming).I can&#x27;t help but wonder what developers are doing in their workplaces. Is SDD actually used as part of the developer toolkit? Am I trying to keep too much control over the code and should I just let the LLM be free within the bounds of a spec to guide it? Or should I abandon SDD and look at another path to ensure that code being produced can at least be maintained long term?

Ask HN: Are you using Spec Driven Development?

Students need compute, please consider helping the next generation.

DigitalOcean Ending Student Credit Promotion

They get lost in SHOWN HN too often. This thread is for them. Unfinished is ok. Just show them.

Ask HN: Show your AI coded games [June 2026]

Ask HN: How to avoid stressing yourself because of a micromanager?

The post with the title:Gauntlet AI will fly you to Austin, train you in AI, give $200k+ job (gauntletai.com)The post doesn’t have any author attributed or comments. Is this something new?

Ad on the HN front page? Gauntlet AI?

I keep pushing the frontier models to the limits and have several projects they still can’t solve, I benchmark new models on. Every new model make it easier to “solve” the even harder problems, but still I have this feeling that they rely 99% on my ideas.
They just don’t get the ideas and I have to hold their hand and help them.Don’t get me wrong, anything that is close to done already they excel at and can combine existing techniques. I’m talking about new ideas models have never seen before.Example I have this hobby project that push what’s possible with route optimization. Yes it’s close to SOTA and way more efficient than all (?) other solutions out there (punnerud.github.io&#x2F;mpee&#x2F;), but I have to hold the model in the hand and brainstorm ideas on how to compress a matrix.And it’s just not a one time thing, happens like 40-50 times in few days.The 1% there is this “new ideas” part. Why can I come up with all these, and not the model?
A really hard reval to create. Now this project is open, later I am thinking about making a frontier project in the same way, keeping it away from the public and using it as a benchmark. It’s that the best way to test for new ideas in models?

Ask HN: Just me feeling that Mythos/Fabel just 1% there?

I submit an extension (an adblocker) to Google Chrome&#x27;s web store.Google keeps rejecting it for dubious reasons. The first rejection was claim it was &quot;spam&quot;. When I appealed, the review came back that it contained &quot;additional functionality&quot; because it uses &quot;modifies network traffic&quot;. Well of course it does! When I asked the reviewer how I could achieve the stated functionality of blocking ads without the use of &quot;declarativeNetRequest&quot; I simply received the same canned response.I submit a totally new update that simplified the code and included comments, and references to other open source projecs that use the exact same mechanisms. Again it was rejected. On this appeal I asked if it could be escalated to a senior reviewer who could possibly reply with more context. Same canned response and rejection.I can&#x27;t help but think Google has some internal policy to give adblockers a runaround. It is obvious to me the stated rejections are not valid (Note that I&#x27;m also not new to this - I have several extensions that have been published for years with thousands of users) but have never encountered such gate-keeping before.It&#x27;s a sad state of affairs even if not totally unpredictable. To be honest I&#x27;m surprised Google hasn&#x27;t made it an official policy to prohibit adblockers. But they haven&#x27;t - obviously there are other adblockers published to the store, and that&#x27;s what makes this so frustrating.Anyway, fellow developers. Anyone run into a similar situation? And how did you resolve it? Thanks!