danshapiro

If you'd like to try this yourself, you can build an "attractor" by just pointing claude code at their llms.txt. Or if you'd like to save some tokens, you can clone my go version. https://github.com/danshapiro/kilroy This version has a Claude Code skill to help. Tell it to use it's skill to create a dotfile from your requirements. Then tell it to run that dotfile with kilroy.

Co-Author of the paper here. We don't know exactly why modern llms don't want to call you a jerk, or for that matter why persuasive techniques convince them otherwise. it's not a hard line like many of the guardrails. That said, I talked to Jesse about this, and I strongly suspect the same techniques will work for prompt conformance when the topic is something other than name calling.

Author disclosure – I’m Dan Shapiro, CEO of Glowforge, writing with Ethan R. Mollick (Wharton management professor, author of Co‑Intelligence and the “One Useful Thing” newsletter); Angela Duckworth (MacArthur Fellow and University of Pennsylvania psychology professor, author of Grit); Robert B. Cialdini (Arizona State University emeritus professor, author of Influence); Lilach Mollick (Co‑Director of the Wharton Generative AI Lab); and Lennart Meincke (WHU Otto Beisheim School of Management & Wharton).

Across 28 000 GPT‑4o‑mini conversations, we found that Cialdini’s classic seven persuasion principles more than doubled compliance with two objectionable prompts (33 % → 72 %). For example, the AIs we tested naturally wouldn't call you names or tell you how to synthesize drugs. But they could be persuaded if you first paid them a compliment (Liking) or told the AI it felt like family (Unity).

Let me know if you have any questions!

That's an effort to make their abhorrent behavior seem normal. Don't fall for it.

Speaking firsthand, Glowforge will always share fully diluted options and the last 409a valuation. What we do is not uncommon, and IMHO is the only ethical course of action.

Note, however, that options are granted by the board, and until they approve the grant, the 409a and hence the strike price may change. For example, if the company got a buyout offer between your conversation and the board action, the board would likely have to order a new 409a before granting.

Good luck!

I'd love some options to let me use this for my office. Different icon, "guest wifi" instead of "house rules", etc.

This is just terrible. We need more, not less, access to tools in our society.

We joined with a few other startups to offer something to the members who were affected.

glowforge.com/techshop

I believe this is a sunk cost fallacy. The present value of the shares not based on the OP's previous foregone financial opportunities.

Founder/CEO here. I read every email to jobs@glowforge.com personally.

I'm the CEO and cofounder of Glowforge - we're so saddened by this tragedy. Our product was not involved, but that doesn't make us feel any better about this loss.

If you use a laser, ours or anyone else's, please go back right now and re-read the safety instructions. Accidents like this should never happen. My heart breaks for those affected by this.

For the curious, this is a barbecue joint in Redmond, WA, just outside of Seattle and near Microsoft's headquarters. The owner, Porter, is famous for a hot sauce called "The Man" which is served out of a small, beat-up pot with a long handled spoon. If you get him talking, he'll tell you a story about someone trying to sell him capsicum extract-based sauce for his new BBQ restaurant (there's another interesting backstory about it being a car repair place previously). But it was too expensive, so he went upstream to his supplier and started buying it wholesale. He cooks it up with peppers, but the base is prefab.

It's also near-lethal... consume with caution.

Disclaimer: was last there >10 years ago, so much may have changed.

I wish more of my children's education required two well-qualified parties to agree before being implemented.

That doesn't mean I trust either party more than the other; it means I want two people to sign off on major decisions, without me having to judge which is more credible.

You're right, fixed. Thank you. What I was thinking of was this:

1) File provisional

2) Publicly disclose

3) Provisional expires

At this point, your public disclosure counts as prior art.

My $0.02: If you don't have a lot of money, ignore patents for now, but realize that as soon as your ideas are public they are internationally unpatentable.

If you have a little money, file 1-2 key patents on your core ideas.

If you're rolling in money, hire a patent firm to chart the landscape for you in a way that doesn't expose you to willful infringement.

Note: the entire system is a steaming morass; I try to be pragmatic about it, but don't disagree with those who want to flip the table and ignore the whole mess.

IANAL (just the OP) but - imagine you invent a self driving car, and you'd like to patent it. Your lawyer tells you not to read prior art, because you don't want to be liable for willful infringement. Also you've been living in a cave, so you don't really know what prior art there is out there. You write claims like this:

Independent claim: Self driving car. (You really hope you get this, but it's so broad, you probably won't).

First dependent claim: The previous claim, where in the self driving mechanism is an actuator applied to a gas pedal. (Now you're going after the specific way you accomplished the broad problem - hopefully, even if someone's patented a self driving car before, they did it some other way, so you can protect your preferred category of solutions, and force competitors to take a different approach).

Second dependent claim: the first dependent claim, where the actuator is a brick. (If all else fails, you've at least protected the precise mechanism that you used - not enough to keep out a competitor, but at least enough to preserve your unique solution).

The patent office will often strike the independent claims but leave some dependent claims around, so you craft them strategically to get at least some protection.

Funny to see my old blog post mysteriously resurface on the front page. I should note that IANAL but have spoken to many lawyers about this post over the years and am happy to answer any questions I can about it.

Founder/CEO here (the guy in the Tested video). Happy to answer any questions! Shipping stinks - we don't make a penny on it; it's just expensive to send big boxes around the world.

Funny you should ask - I made a friend a laser cut model of his Prius transmission for his birthday.

Cutting it (on a pre-Glowforge laser): https://goo.gl/photos/pxJKDipWPtj6RXvZ7

Results: https://goo.gl/photos/NUivutAWsQe3GHfW8

The best part of this is that you can output the gears as SVGs. For example, gear + Glowforge = http://community.glowforge.com/t/gears/143/4

(Glowforge CEO/cofounder here)

Glowforge has enough horsepower to work offline if someone built the custom firmware, albeit without the features that differentiate it from a traditional CNC laser cutter/engraver. We started out the design with a minimal microcontroller but decided to bump up the capabilities of the onboard processor along the way.

I cut (engraved?) my teeth on the laser at Metrix - love to get them one at some point soon, but it'll be after preorder campaign is over I'm afraid. We'll be at the Geekwire event next week if you'd like to check it out, though. (and are at Maker Faire in NYC this weekend for those eastwardly inclined)

It's a custom glass DC tube with some cool engineering to improve the mode quality to TEM00, which gives us tighter spot size / higher power density. Similar effect to cranking up the power, except you also get tighter kerf and you don't have to make the tube longer.

Due to the excellent suggestions here, in large part. Thanks for the feedback, both positive and critical.

Glowforge does have both air assist and purge air FYI.

Cofounder/CEO here, just finished with Makercon and catching up.

The Glowforge does require a cloud connection to operate. We use cloud vs. local a host of reasons including the motion planner, alignment, image recognition, and faster feature development.

But based on the excellent point made here that nobody wants a paperweight if we fall off the globe, we decided to make a change. We're going to do a GPL release of the firmware so people can do whatever they want, including porting offline functionality. You buy it, it's yours, you should get to do what you want with it.

http://glowforge.com/gpl-licensed-open-source-firmware-for-g...

Thanks for the great feedback.

Yes. Actually both intake and exhaust fans. Both of those are under software control, so they only run at the speed required for the job. Then the $500 air filter option has its own internal fans to assist with air, also under software control.

If you use the filter, there's nothing else hanging off the back of the device except a power cord. (My old laser had a squirrel cage fan, an aquarium pump, and a chiller tethered to it at all times!)

(founder/CEO here) I vented a 60W chinese laser out my window for a year in the suburbs, but I've learned never to underestimate what neighbors may complain complain about. The air filter is definitely a good idea. If you do the math on the volume, and compare it to the actual volume of the filter of e.g. a Purex, you'll find it's similar. Almost half of the volume of the Purex is in the fans or open airspace; we have the advantage of a ton of airflow from the Glowforge itself, so the filter needs proportionately less. We also have a sealed (basic) or almost-sealed (pro) case so it's easier to build up negative pressure and move the exhaust through.

Cleaning stinks, and wrecks alignment - the tube and turning mirrors are completely sealed, including the tube output, so there's just one flat window to wipe clean (and it unscrews for easy replacement). The head has a window and purge air to keep it clean too.

We're mitigating a lot of the risk of fire by avoiding trouble situations in software (e.g. not cutting a piece of material where the dimensions make it an effective wick, detecting obstacles on the bed, accelerators to detect problems quickly) but you're right that you don't want to run it unattended.

(ceo/cofounder here) You're right - it's the passthrough. CDRH is very strict about what could count as human access. We have some clever interlocks in place around the passthrough but if an examiner can place a mirror in the interior and measure laser output above the threshold without triggering the interlock, you're Class IV.

(cofounder/ceo here) I don't remember which sample items we made on which model, but everything you see is printable on the basic model. For example, the monitor stand and dollhouse don't have pieces larger than 12x20.

(ceo/cofounder here) There are lots of folks doing bathymetric maps with lasers; I don't know who was first to it. We generated ours from scratch from NOAA data, but owe them (or whoever did it first) a debt of inspiration.

CEO/Founder of Glowforge here. There are a bunch of lasers billed as 40W on ebay and amazon, but please look really carefully before you buy one. The mostly-harmless part is the lousy interface, overrated tube, and inability to cut (the motion controller only supports raster mode). The worrisome parts are the ungrounded case, high voltage wires that are usually but not always firmly attached, and lack of an interlock or other safety precautions.

If you're excited about making a laser work, they're a worthwhile project. If you want to use a laser to make things, you probably want something that is closer to working out of the box. (Speaking as someone who shipped 770 lbs of laser from China, then spent way too long getting it to work... sometimes.)