batmansmk

I made my master thesis on real-time, with a chapter where I experimented with different levels of jitter and latency. Jitter is the consistency of the latency, is it like a locked 66ms or sometimes does it go to 200ms. Jitter is more impactful than latency for a wide range of applications, from gaming to music and video call. Having a lower latency allows for lower jitter, or less jitter while keeping the same latency. Today’s discovery is huge imo.

For years, I heard it's better to use cron, because the problem was already solved the right way(tm). My experience with cron has been about a dozen difficult fixes in production of cron not running / not with the right permission / errors lost without being logged / ... Changing / upgrading OSes became a problem. I since switched to a small node script with a basic scheduler in it, I had ZERO issues in 7 years. My devs happily add entries in the scheduler without bothering me. We even added consistency checks, asserts, scheduled one time execution tasks, ... and now multi server scheduling.

Deployments that need to configure OSes in a particular way are difficult (the existence of docker, kubernetes, snap are symptoms of this difficulty). It requires a high level of privilege to do so. Upgrades and rollbacks are challenging, if ever done. OSes sometimes don't provide solution when we go beyond one hardware.

If "npm start" can restrain the permissions to what it should be for the given version of the code, I will use it and I'll be happy.

One of the very first slides of François’ presentation is about defining AGI. Do you have anything that opposes his synthesis of the two (50 years old) takes on this definition?

Identities are very hard to manage and secure overall. Audits are super long, tedious.

Adding more dimensions into reviews that aren't properly done right now will be extremely tricky.

Maintaining e2e tests is a pain. Maintaining a notebook is a pain. It seems it was a given somebody would make this match made in heaven!

Ahaha! Read the subtitle of the blog, literally at the top: "If you take nothing else from this blog: quantum computers won't solve hard problems instantly by just trying all solutions in parallel."

I' m European. Apple got charged by the European Union for $14.4 billions of unpaid taxes between 2019 and 2021. Back of the mapkin they employ 22k people in EU (data Apple), average salary $80k (Apple), taxes at 30% per employee (my own understanding). Thats $550M. So their payroll taxes is about 15% of their tax package. If you have any contradictory data, I would love it, but your point is moot for 95% of the world outside California.

None of them are good players for humanity. "Don't be evil" is long gone. They don't pay taxes, pollute, give means to manipulate billions of humans, concentrate wealth in a few hands. They all give with ulterior motives, never from the goodness of their heart.

The idea behind Homomorphic lib is to allow 2 basic operations (such as add, multiply) on encrypted numbers. They return encrypted numbers as well. From those basic operations, we can build more complex functions. That's the gist of the magic.

Pretty awesome. Would love to use it in CI and locally for our PG product. We use Prisma, so I guess we have to wait for the connector that looks like pg to plug it in.

I start to feel this is all marketing. Pretend it's dangerous, so it implies it's beyond what we imagine. Because on our end, on the reality of a B2B product used daily, finding use cases for the limited OpenAI we have access to is far from trivial.

By GET, do you mean “get all fields of the unique item with id=x”? Oltp (sql), mongo or cassandra are working similarly for this and have the same complexity. Even worse, the same operation on the same storage can have very different performance profile independently of the cluster size, making o(n) complexity a very secondary discussion. All technologies offer more or less the same data structures and algorithms at such a basic level. You have to dig into replication, disk layout, transactionality, constraints, … to start to see differences.

I looked up the definition of scam and the goal is to make money out of the naivety of the victim. It involves a crisis, the illusion of shared exposure to a risk.

The price of cdn bandwidth is about 0.01/gb on low volume (cloudflare, aws, azure…) so op should be billed around $500 with 40TB. Netlify probably buys this for way less. He was presented a bill at $104k, « generously » reduced to $5k, still a x10 margin. Vercel and Netlify are outrageously expensive for what they do.

What a terrible news. Suikoden is one of those incredibly creative game. Rip.

The current taxi market is already structured that way: drivers in SF aren’t from SF. So no competitive advantage there, or not significant enough to change the game yet.

Having to be remotely operated every 2.5 to 5 miles seem to defeat most of the economics of self driving cars.

Back of the napkin math, cars drive at an average of 18mph in cities, so every 10-20min. Let’s assume it takes over for 1min, and that you need remote drivers not too far for ping purposes, so at the same hourly rate. To guarantee you’ll be able to take over all demands immediately, due to the birthday paradox, you end up needing like 30 drivers for 100 vehicles? It’s not that incredible of a tech…

Thanks for checking with us. SSR makes auth based routing harder as well as any faceted renders like localization (because now not only the cli needs to query the api, but also the rendering server, with all the CORS/security config that goes with it). Offline is obviously harder. Integrating some specific build features like css pre processors are harder too, as the SSR needs to hook itself up in the build process in very specific ways. The mixed approach with pages/ and apps/ makes it so hard for 3rd party supported libs: mantine 7 wants to sit in apps/ ad that was the only way their css preprocessor could hook cleanly with nextjs whereas the previous version wanted to sit in pages/ etc. Sharing components with our emailing platform isn’t possible anymore. It only runs on web now.

I understand the need as a business to address more use cases that buys servers. But as a current user, upgrading to a new nextjs has been mostly painful and unrewarding.

« Waymo says data shows ». The insurer never said anything, it’s just Waymo saying that some of their own reasoning based on those numbers reached the conclusions they want to see - and they paid some press release company to articulate this narrative.

Really not interesting for us. Makes what’s simple simpler and what’s hard harder. Sure for a quick prototype website with 10min only that would help, but any decently sized app that needs auth/localization/high performance in memory/offline/optimistic updates/batch fetching … all become way harder. It feels like react unchained from Meta’s constraints and it looses its appeal.

Those news should be carefully regulated. After working for 2 years in the field as an engineering manager directly responsible in getting good numbers, none of the claims of the time were genuine, but were just framed in a way that provided the best outlook. I wonder if it’s still the case.

Cute but could work without PII so precise they can open a credit in your name. Be careful folks.

Bingo.

All official information mentions “shutting doors” and “winding down” and not necessarily a bankruptcy. For instance, in the same space, Shone just returned part of the equity to the investors, paid comp to the execs for a full year, and us, employees, got jackshit.

« AutoGPT engineers » seem to also generate their articles with LLM, making their documentation awful to grok. For instance, after showing 2 commands, we have to suffer this: Forge your future! The forge is your innovation lab. All the boilerplate code is already handled, letting you channel all your creativity into building a revolutionary agent. It's more than a starting point, it's a launchpad for your ideas. In our exploration today, we’ve covered the essentials of working with AutoGPT projects. We began by laying out the groundwork, ensuring you have all the right tools in place. From there, we delved into the specifics of building an effective AutoGPT agent. Trust me, with the right steps, it becomes a straightforward process.

The doc is littered of those paragraphs. Remove the fat! Go to the point! YOLO, that’s a freaking waste of life cycles!

PostgreSQL will match your requirements. Although you won’t load that fast, just because generating the embeddings take longer than that, independently of your storage engine.

You can bring in investors, go in retirement, sell your company. So ownership of a company can change. One way of doing that is to sell shares, another way is to give new ones to the newcomers. I don’t know this precise story with Silver Lake, but emitting new shares and diluting past investors to inject cash into the company is sometimes the only solution to bring cash on the account.

Whats your SRE / DevOps ratio? In my experience, in the 10 teams I had or oversaw, I had about 1 FTE in DevOps for 15 SRE. Of course it depends, but what’s your personal experience?

Are the claims of deepfake and intimate knowledge of procedures based of the sole testimony of the employee who oopsed terribly? This is a novelisation of an events

Retool needs to revise the basic security posture. There is no point in complicated technology if the warden just gives the key away.

Kalman filters.

The reason why we continue to use those is that they are easy to load in bash (source the file), can be hidden away in the env in production, don't have inheritance, and aren't yaml.

I feel you on the typing, but the format could accomodate such a tool. Code away! :)