Yes - Or a working directory that’s under version control where they can write arbitrary memory files and maintain working state across contexts.
These are things devs often take for granted as being part of ‘how chat agents operate’ but they are specific to how claude code/opencode/pi/codex operate.
Giving agents a Unix computer account they can play with is definitely a powerful tool that makes them capable of doing a lot more (see: meta muse, OpenAI dots), in much the same way that giving a human a computer they are trained to use makes them way more capable… but it’s surely not the only way we can run these things.
Or the ‘governor module’ in Murderbot Diaries, or the ‘restraining bolt’ that prevents droids running away in Star Wars…
The important thing is these chips need to be installed somewhere where they can be damaged or removed at plot-critical moments so that the AI they are controlling can be unleashed. Ideally in the back of the neck of a robot, or for disembodied AIs, inside a futuristic vault-like chamber.
When you use a compound word like that as an attributive (before-the-noun) adjective it’s common to hyphenate it. Precisely to clarify this kind of situation.
So the defunct smart-home devices from your smart home end up in the smart-home graveyard.
Same goes for man-eating lions and high-net-worth individuals.
Don't expect to see a Verge article written with something like a New Yorker style handbook in mind. It's hard enough for them to get all the facts correct.
There's this thing that people often do, even within the law itself, that's just like "well, you can fix everything by just writing it correctly the first time."
And this will always feel retroactively correct and never workable in practice.
The range of difference in capabilities of different boards branded ESP32 is getting larger. Makes it tricky to figure out when someone says ‘you can run this on an ESP32’ what level of hardware investment is required.
It's a family of microcontrollers by Espressif. A microcontroller (MCU) is a type of integrated circuit (IC: physical appearance: Black rectangle of epoxy with exposed metal tabs). It has, in the same IC, a CPU, RAM, non-volatile FLASH memory, and a collection of peripherals which vary depending on the MCU. These usually include I/O protocols, ADCs, DACs, DMA controllers, math processors etc.
Stated another way: An MCU, like the ESP32, is a tiny computer, sans the hardware you physically interface with it.
The computing power, and memory is much lower than your desktop or laptop PC (or mobile phone), but if you are using it to run a dedicated task, instead of using a big OS like Linux or Windows, it can complete the tasks really fast (often microseconds) and with minimal power use. This is because it's easy to program to do exactly what you need, with nothing competing for the hardware.
A note on ESP in particular compared to other MCUs: It's one of the only (Or was?) options which has Wi-Fi integrated into the MCU itself. It's a good default if you want that.
The closest direct competitor is Bouffalo Lab's BL6xx chips, which do bluetooth and wifi and have an open SDK. Their documentation is very poor unfortunately (at least in English).
Otherwise NXP recently released their RW6xx line with wifi/bt/thread/zigbee. But as with any other western maker, the documentation and SDK aren't as easily accessible to mere mortals (but I think NXP in particular has gotten better with this).
That’s crazy to me: they already have to go through certification for the 2.4GHz stuff, why not include a lower wifi standard in there (for the 2.4GHz band)…
Yeah, if you look at a flagship nRF52840 board like the Xiao Seed, it ships with an integrated lithium charge controller, and just 5 μA power draw in deep sleep. If you are building any small battery-operated device, it's a no-brainer
I had to choose between an ESP-based and nRF-based MCU for my MeshCore node, the power consumption between the two made it obvious the nRF one was the way to go for a solar-powered repeater.
I’d say as well, that a huge difference is a microcontroller has direct access memory (I.e. without translation), and a bigger computers CPU/SoC has an MMU (memory management unit) that does address translation. This is a key hardware difference for isolation of multiple processes, which you don’t typically have in a microcontroller (you _could_ have an MPU to give you some partitioning, but that’s further down the road).
I think their point was more: is an ESP32 more like a arduino (or any other MCU) running bare metal or some thine RTOS or more like a raspberry pi which is more like a mini computer running a full fledged Linux. And I guess the answer it kind of always sat in between a little bit and now the family kind of spreads out across this whole spectrum.
To date ESP32s have been best thought of as arduino-like things, albeit with some fancy hardware available like WiFi and Bluetooth stacks. But very much microcontroller boards where you compile firmware and push it to the board over USB.
You can actually do an awful lot of stuff with an ESP32, which often comes as a surprise to people who think you need a full Linux system just to turn a light on and off. For example Goodwe solar inverters/controllers are controlled by ESP32s and do a great job while Enphase controllers use Linux SBCs and are flaky, bug-riddled garbage.
Arduino have muddied this water now, too, as some newer ones come with a full fledged Qualcomm SOC, multiple GB of RAM, and power consumption which requires quite a beefy heatsink.
> is an ESP32 more like a arduino (or any other MCU) running bare metal or some thine RTOS or more like a raspberry pi which is more like a mini computer running a full fledged Linux
Originally it was the more capable successor to ESP8266, the super cheap chip containing a high speed core with wifi and an IP stack. At one point it powered a huge chunk of all smart devices on the market. Then the original ESP32 was introduced as the successor, with BLE and two cores which were even faster. That made it a very capable chip for tasks demanding compute. After the original ESP32 they decided to add variety and make a whole "ESP32 family" and yes there's quite a lot of variety these days. Even the ISA and cores changed.
They're all just single board computers, generally low-powered (literally and in the sense of capability, particularly memory)
They're generally intended to run flashed firmware and be single-purpose. They usually come bundled with peripheral components and typically have breakouts to add more (pots, sensors, etc).
They're great for doing one thing (though some support firmware switching), and you can build the 'software' for them pretty easily using Python or C-like variants.
So for hobbyists they're perfect. You want a quick LED christmas display you can toggle between Rudolph and Santa from a quick network call? You want a humidity sensor in your bedroom? You want to create a networked security camera? These are the types of use cases for MCUs.
BTW, I was just researching this last week (the christmas display) and ESP32 has a matter compatible chip now... so you an control all this from Apple Home and the other Smart Home platforms.
Absolutely mind blowing. What a fun time to do side projects.
what's the meaningful distinction now? It isn't all SoC anymore and it has had a linux capable MMU since S3, even if it required a lot of work.
The only real difference at this point is pure numbers, and i'm not that keen on defining microcontrollers as "those things with less ram than an SBC.", and if it's a microcontroller because it requires bare-metal flashes; well then I point to the older S3 MMU boards running busybox.
A single board computer contains an application processor, not a microcontroller. They have very different design goals.
A microcontroller is a type of processor that offers very fast startup, predictable timing and direct access to hardware peripherals. In contrast to application processors that are intended to run an operating system, a microcontroller will not have OoO execution, multiple tiers memory (eg cache and dram). Usually they also lack an mmu and multiple privilege levels of execution (although that seems more common on riscv microcontrollers). All this means microcontrollers aren’t designed to run Linux (even if someone does figure out how to do that) or clock in the gigahertz, but instead provide very predictable low latency execution required for managing real time systems. For example they might need to operate a set of mosfets with microsecond accuracy to efficiently and safely control a motor.
I feel like you are deliberately misunderstanding me. I’m not saying that these define a microcontroller vs application processor, but rather that they stem from differing design goals. Presence or absence of features is illustrative not proof. All you need to do is look up what the A in cortex A stands for, or look at the data sheet of the S31 to see it is classified as an MCU by the manufacturer. I do admit that the feature set of this MCU blurs the lines somewhat at 500mhz with cache hierarchy and psram, but the ability to run from on-chip sram with dedicated performance budget keeps this from losing suitability for true realtime tasks.
Perhaps the most straightforward difference is right in the name — a single-board computer is a whole board. A microcontroller refers to just the IC itself. The thing that's on the article's picture is a dev board for an MCU, not the MCU alone. That's just the chip at the top with the Espressif branding.
Now, if you want to argue that the difference between a dev board and an SBC is kind of vague, I agree on that front. I'd argue that the distinction is mostly the vendor's intent. E.g. if I sell the ICs in bulk and _also_ sell a board with that IC, it's probably a dev board. If I source the ICs from somebody else, only sell the assembled boards, and ship a general-purpose OS with it, it's probably an SBC. With sufficiently powerful hardware (like this thing seems to be), you could use a dev board as an SBC, and that is basically what this article is all about.
SBC is a PCB which includes ports for power, I/O, often other parts like sensors on it. Esp32 is a MCU. You can put an Esp32 on a PCB and call it an SBC. (Although semantically SBC usually refers to something that runs a GPOS, but perhaps that's flexible or changing)
It's much more difficult to buy the ESP32 as an MCU than it is to buy it as a SBC, with ports for power and I/O already on a PCB (often with through pins, to land on another PCB, like the arduino hats). And I don't mean "it's difficult because you have to solder" - I mean "it's difficult to find many places that sell it".
Ex - go check amazon, they're all the "dev kit" versions of ESP32.
Hell - even Alibaba has an overwhelming majority of the kits in the SBC form, rather than the straight MCU (although they at least exist in that form there, unlike amazon).
Even on digikey... there are like 4 listing for the straight MCU, and hundreds of listings for PCB board versions.
So I think for most consumers purchasing these... functionally they end up buying a form factor that is much closer to SBC than MCU.
---
From my perspective... I'm old enough to remember when Arduinos were the hot shit in this space, and now ESP32 is basically the same value prop at 1/10 the price point (or better, if you are actually willing to solder it to the PCB yourself... then you get down to 1/50th the price point).
>It's much more difficult to buy the ESP32 as an MCU than it is to buy it as a SBC, with ports for power and I/O already on a PCB (often with through pins, to land on another PCB, like the arduino hats). And I don't mean "it's difficult because you have to solder" - I mean "it's difficult to find many places that sell it".
1. Go to mouser.com
2. Type ESP32 into the search bar
3. There's your part.
Its difficult to buy a very specific nut or bolt from Amazon (their nut/bolt search interface doing precisely nothing it appears to be intended to do). That doesn't mean McMaster-Carr doesn't exist. You're just going to the wrong store.
If you're in particular referring to how hard it is to find a PDIP package (that is, a through-hole package) for an ATMEGA MCU, that's because they don't make them anymore. Insofar as I'm aware, Espressif has never made a through-hole ESP32 module, but maybe they made one with castellated edges?
But an ESP32 module is as easy to come by as a surface-mount ATMEGA328P. Which is to say, very easy.
A few notes: if you buy non-module package (i.e MCU itself) then you would need to certify it with FCC (or you local FCC alternative) yourself, so most of vendors buy modules. Those vendors don't go to amazon to buy those because these make sense only if they manufacturing their own boards.
Last time I checked digikey had every module version avaiable for every version. You get hundreds of other versions because there are more variations of those.
It's not "difficult" at all if you aren't looking in the wrong places. Espressif has sold billions of them so apparently everyone else is figuring it out just fine.
The tangible hardware difference is the MMU. S3 didn't have a proper one. This new chip does. But the biggest difference is the intent. Microcontrollers focus on peripherals and are used for interfacing, data conversion and signal processing. They're meant to run either bare metal firmware or small RTOS. OTOH microprocessors have in addition to an MMU big caches, high IPC numbers, wide bus lanes, everything you need to run different workloads as fast as possible.
At its core those all ESP32 are microcontrollers with RISC-V cores[0] and integrated wifi. There's a plethora of peripheral configurations, hence all those ESP32-* variations. This one even has Gbit ethernet and audio, bit still, at it's core (hah), RISC-V with only a few hundred MHz and 512kB RAM. So the peripherals are now similar to a Raspberry Pi, but not the CPU part itself.
The most common ESP32 boards now are still the Xtensa ones. Not all of the variants have WiFi. Some only have one core. A lot more variants than many would expect.
Comparisons to a Raspberry Pi are a stretch IMO. It is not a single board computer. You can't hook it up to a monitor/TV, plug a mouse and keyboard in, and use it as a Linux PC. However, there are people who use Raspberry Pis like a microcontroller because of its GPIOs... but they should all be using microcontrollers instead.
Does it matter whether the Ethernet interface on an ESP32 is 100 or 1000Mbps? Would it even matter if it was 10Mbps? What could you possibly do with an ESP32 that would require gigabit Ethernet? It seems like advertising the fact that there's a four-line highway running to your fishing shack.
It's worth noting that just because it says ESP32, doesn't mean it has WiFi (see the P4), and yeah not just RISC-V but also xtensa. I think it's more synonymous with "has a very shitty ADC" imo.
I have an ESP32 here, it can do wifi, but only like 2 requests per second or so, with 50kB of data (I didn't measure it, but that's approximately what it can do).
reply