Last sentence completely undercuts the other sentiments you shared in your comment… probably best to cut stuff like that out in the future IMHO, even if it’s how you feel.
This doesn’t seem plausible. Leading edge lithography requires you to be at (or beyond) the cutting edge in many realms - even with a breakthrough in one realm, I don’t understand how a startup could expect to catch up to ASML across the board in a few years.
Their website is light on technical details and heavy on nationalistic fluff, which does not lend much confidence.
It is actually relatively easy to make a lithography machine that can etch features beyond what EUV can do. You simply use an electron scanning beam rather than photons.
It's what the industry uses to create the masks used in lithography machines, but it could just as easily be used to make the actual chip. The problem is that it doesn't scale, at all. A scanning process is way too slow to be useful in mass production.
Thus you should always be skeptical when someone says they've built a machine that beats ASML's machines, because that's actually the easy part. The hard part is scaling it up.
Interesting! Makes me think of old 1990s X-Files episodes with chips under a microscope “smaller than we can produce”.
I wonder if the government makes small batches of bespoke chips that are super miniature based on non scalable processes, and how far back in time would they have been able to develop 1nm chips for example?
The node sizes have become more of marketing term, so it's more useful to look at the half-pitch resolution when doing comparisons. In 2007 researchers demonstrated they could reach a 15nm half-pitch using an electron scanning beam. [1] Whereas ASML reliably achieved this resolution using EUV around 2017.
Thus in the early 2000s you would be about 10 years ahead using electron scanning beam lithography. However that assumes you have all the tooling and transistor designs to actually create a working chip at that resolution. Showing you can etch a feature at nanometer scale is one thing, actually using it to create a working chip is a whole other ball game.
Funnily enough Asianometry just did a video on tsmcs new masks and how the machines involved WERE particularly hard to develop, "Multi-Beam Mask Writer" that uses hundreds of thousands of electron beams (after splitting) to accomplish its task.
> investments from the Central Intelligence Agency-backed nonprofit firm In-Q-Tel
The CIA has stolen trade secrets in the past and the only thing that stopped them in recent history is their own policies. The CIA has a new director that has been violating international law more openly than ever.
Exactly. And it should. The "CHIPS Act" should be thought of as a perpetual blank cheque to whoever can build the components necessary to build war machines completely with North American components (primarily USA components but Canada will have some impact)
I don't know much about the technical details of lithography, but I do know that EUV lithography is very new tech that has been in production for less than 10 years and the current machines are basically rube goldberg devices. Given my lack of technical knowledge here, I can't say whether or nor this startup in particular is legit, but it does seem very much like the type of thing that could be disrupted by someone who comes up with a new and massively simplified design.
EUV machines were in development for nearly 20 years before they could reach actual chip production.
The secret sauce was not getting it to work, but getting it to work stable enough such that it can be sustained millions of times per second. I am sure there were other huge challenges in bringing to market though, I am not an expert on this either.
Yeah, that's kind of my point. The design is so complicated that the hard part is actually getting it to work reliably in production. So it could just be that the current way is the only fundamental design that works and there is no radically simpler way to make EUV lithography work, but 99% of the time there is.
I don’t think that is a practical framework for situations where people aren’t already very closely aligned.
What happens when a few people are very vocal (and firm) in opposition to basically every change? Having dissenting views is valuable, but not when they have veto power.
Additionally, I think that framework is vulnerable to what I refer to as “death by yes but” - when everyone is just piling on amendments and precursor conditions, oftentimes conflicting, that result in a decision taking months (maybe even years) to make or scuttle.
I’m basing these comments out of experience - one example is a workgroup/committee operating under a similar model that was completely unable to do anything due to decision paralysis. The committee grew significantly more effective when we reformed the decision making process to have a small group of owners to handle pitching and (potentially) implementing the decision, then had approval be a simple yes/no majority vote.
>> You could learn from consent based decision making, a hallmark of sociocratic worker coops that is underrated and can be applied elsewhere.
>> Hierarchy and coercion isn't necessary for avoiding decision paralysis in organizations.
> I don’t think that is a practical framework for situations where people aren’t already very closely aligned.
Putting aside the concept of Sociocracy for the purpose of discussing engineering team leadership philosophies, one which I have observed to be very effective when working with experts is Servant Leadership[0]. From the Wiki page:
A servant leader shares power, puts the needs of the
employees first and helps people develop and perform as
highly as possible. Instead of the people working to serve
the leader, the leader exists to serve the people.
While Servant Leadership[0] might initially raise concerns resembling the problems you rightly identify with a sociocratic approach, it has the benefits of peer collaboration combined with accountability of the decisions made by leadership.
Yes it only works when participants have a shared aim
In full sociocracy, it...
> honors the circle’s ability to freely make decisions within its domain. This is key for the organization to remain effective. But it comes at the cost of members not having “consent rights” to every decision the organization makes. Each member will only have those rights for the circles they are a part of.
So it's not necessary to allow people outside that working group to veto or require compulsory followup through objection
There's no perfect solution to organization, everything is a tradeoff. I've also been part of working groups (made of people whose job description is to manage the scope of changes they're proposing) where everyone and the workers impacted by the decisions are in agreement and no impact on cost etc., but the exec decides no change can be made due to personal preference despite disastrous consequence. Or where an exec who abstains from checking in on a working group's efforts nonetheless counters it with shifting and contradictory demands whenever it comes time to take action, requiring going back to the drawing board repeatedly until people simply give up or leave the organization. Or where the exec asks for more data for a proposal, and then doesn't evaluate the data once gathered, leaving no recourse but to give up or leave the organization. We all have stories like this. Hierarchical organizations are also susceptible to paralysis.
In CO we have automatic traffic cameras, and to my knowledge they just mail you the ticket, which is usually only a fine (and no license points). Its one of those “automatic plea” tickets where if you fight it, you fight (and risk conviction on) the actual offense, while if you just pay the ticket it will automatically get downgraded to a less serious offense (IE parking outside the lines).
I suspect decreasing tips also has a lot to do with tip exhaustion, I think it's becoming increasingly acceptable (at least in my circles) to hit the "custom" option and put in a smaller-that-suggested tip (or even nothing at all).
I've stopped tipping through the POS devices entirely. Now when I tip, it's in cash and an amount I've decided on, not an amount the business is trying to talk me into. I also have reverted to only tipping for traditionally tipped stuff, not the everyone and their dog who asks.
Ironically, those stupid tip kiosks have cost severs tip amountage with me
I used to drink a fair bit, so at the end of a tab, I did not want to do arithmetic and would just put an “easy” number that looked good. This could often be upwards of 35%
But if you stick a touch screen in my face with even easier, one-two taps on a preselected amount, I’ll go with that.
Also I’ve accidentally no tipped a few times with them because I get frustrated with the incessant screens asking about receipts, emails, etc. that I just click past the tip screen.
I'm making the argument that I wouldn't rely on an API endpoint to serve up the secrets that enable my application to work. Imagine a network outage or endpoint failure when the app just happens to be redeployed.
I think it depends on the API - we do this with AWS Secret Managers. I haven't seen it fail but if did it would only effect new instances coming into service so I think we'd have to be pretty unlucky for it to have a noticeable impact.
It's true that anything that can go wrong will go wrong but I wouldn't use that as a maxim to direct designs - risk is one trade-off and it's significance varies.
Nah, we’ve been doing this long enough that handling network failure is just the default assumption now, we should be designing and coding for it, by default.
