I'm using this model for my first python project, coding using opencode along with devstral and Mistral Large 3. I know it's not as capable as other, more expensive models, but working with it this way is teaching me python. More directly to your point though, the speech to text model is really good.
It's funny because I just took a break from it to read some hn and found this post.
> Yes, MCP eats up context windows, but agents can also be smarter about how they load the MCP context in the first place, using similar strategy to skills.
I have been keeping an eye on MCP context usage with Claude Code's /context command.
When I ran it a couple months ago, supabase used 13.2k tokens all the time, with the search_docs tool using 8k! So, I disabled that tool in my config.
I just ran /context now, and when not being used it uses only ~300 tokens.
I have a question. Does anyone know a good way to benchmark actual MCP context usage in Claude Code now? I just tried a few different things and none of them worked.
This was great info, but as someone who is concerned about passing great info on to others about my own products, I am suddenly less worried about my posts reading like LLM-speak. This post did very well. I should probably stop overthinking it, and learn from this post about balance.
My weekend side project just took over my life. It needs "actual marketing" expertise. As in, I know how to set up search console, semrush, and I know decent SEO concepts to grow organic SERP performance. I am coming up the charts there.
However, I have friendly investor interest. The only place I can imagine spending money on this project is on Google Ads. I have no real idea of how to create and manage Google Ads these days. So, who do I hire? Does anyone have any recs for what I should do? Is there a service, or a go-to consultancy with a small minimum spend requirement?
"Investors in Thompson's venture accused him of cheating them out of promised proceeds and after years on the run he was jailed in 2015 on a criminal contempt charge.
But last year, the judge agreed to end Thompson's civil contempt sentence, arguing that he was unlikely to ever offer an answer, according to CBS News."
It's confusing because it looks like he was convicted of both
U.S. District Judge Algenon Marbley agreed Friday to end Tommy Thompson’s sentence on the civil contempt charge, saying he “no longer is convinced that further incarceration is likely to coerce compliance.” However, he also ordered that the research scientist immediately start serving a two-year sentence he received for a related criminal contempt charge, a term that was delayed when the civil contempt term was imposed.
> The last time I saw this story, I learned that he was actually jailed for defrauding investors. Was that not the case?
> Please use it on this story. Pure garbage.
The third sentence of the story is "Investors in Thompson's venture accused him of cheating them out of promised proceeds and after years on the run he was jailed in 2015 on a criminal contempt charge."
You could call the title clickbait, I guess? (It seems reasonable enough to me.) But I don't understand your objection to the story itself. It makes it clear that the case was about defrauding investors.
Fair. My objection may be founded in other framings of this story where it was implied that "spunky explorer did not tell the big bad goverment where the gold was, so they put him in jail."
Honestly, the headline does seem to imply that, no?
> There's absolutely a time and place for wireless headphones and I probably use them more at this point.
I've had a pair of Sony MD-7506s ever since I assisted in an analog recording studio, years ago. 7506s were the standard then, not sure about now. I am on my 5th set of cushion replacements.
A few years ago I bought a high-ish quality bluetooth adapter I can put in my pocket and walk around. The one I bought is the Boltune BT-BA001. Looks like they are only $30 on eBay now, but there may be better ones out there. That little box is probably my most used piece of inexpensive electronics.
> With all the technology advancement and improvement with access to information in the last 30 years, why does it feel that all of this culminates to more disinformation, more pain, and less understanding?
One of the original adages in technology is:
garbage in, garbage out.
The more technology ate the world, the bigger a problem that became.
If you don't pay for a domain name you could lose it too.
If I signed up for a free social media account hosted by another company and neither logged in or posted on it for a year then it got autodeleted for inactivity, I wouldn't really feel I had a particularly strong claim to it.
One thing that makes handle markets uncomfortable is that social media identifiers sit in a strange space between identity and platform resource.
Domain names are usually treated as leased assets with a clear renewal cycle. Social media handles, on the other hand, often feel more like identity markers, especially when someone has used them for years.
When platforms reclaim dormant handles and then auction them, the model shifts from “resource management” to “asset monetization”. That changes user expectations quite a bit.
If a platform wants to recycle dormant identifiers, a transparent policy with predictable timelines and clear notices would probably feel more legitimate than quietly moving them into a marketplace.
I just launched a free wysiwyg markdown editor. It currently uses only IndexedDB for storage making it as private as possible. The only network calls are polling for the "click to update" toast, and the feedback form.
I was sick of getting cross-eyed when looking at tables in raw markdown and was just running it locally. This weekend I realized it might be useful for others.
The goal was simple as possible UX. Open url, drag and drop or paste into wysiwyg -> very readable and editable markdown. No sign up, no tracking, no fuss.
Of note, if you copy from the richtext mode, it copies raw markdown. The inverse is done with paste.
Based on feedback, I am working on very optional cloud-sync for as cheap as I can make it.
https://md-edit.com now has an MCP server. If you choose to sync with my cloud service, you can ask Claude.ai & Cowork, Cursor, Windsurf, etc to all work on the same set of docs. The webapp has realtime, multi-device sync, version history, diffs, and so much more.
This has taken over my life. I took a week off of work to finish it.
This project was based on dogfooding, and every feature is something I wanted and now use daily. This is the most fun I have had in years.
That webapp I made in a week, is now indispensable to me as the key to human in the loop AI work. Truly, what a time to be alive!
I would like to add that I have also added a Google Drive sync option. It does require creating an MD-Edit account for operational reasons, but I never see your document data in that scenario.
Free multi-device sync is now enabled with email, github, and google OAuth. Of course, privacy is not guaranteed in that case. However, people wanted it for convenience.
I was really trying my best for friction-less UX on this project. I would appreciate any feedback on how I did, either by comment or the feedback button.
reply