Here is a never ending counting of the primes... (except that it will end-up in some arithmetic overflow error after a while).
For less overhead, I use plain PDFTeX. Launching the run in a Terminal, the page numbers correspond to how many primes have been computed so far.
update: the algorithm was a bit faulty, although it computed correctly the primes. The tested integer n had to receive from time
to time a +2 kick that was missing, so that the amount of divisions
done slowly drifted above the necessary. Furthermore my imprudent use
of \1,\2, .. gave unwanted space tokens in the output (this was
very silly on my part, but I have learned my lesson).
..........(terminal output until ^C interrupt)...........
[150451] [150452] [150453] [150454] [150455] [150456] [150457] [150458]
[150459] [150460] [150461] [150462] [150463] [150464] [150465] [150466]
[150467] [150468] [150469] [150470] [150471]^C
! Interruption.
<to be read again>
\TestDiv
\TestDiv ...expandafter \expandafter \TestDiv \fi
\fi
\ToLogIfPrime ->\e \1 \TestDiv
\ifnum \e >\k \write \mone {\the \n }\the \n ...
\CheckNs ...m \j <\k \advance \n \2 \ToLogIfPrime
\advance \j \1 \expandafte...
\FindPrimes ->\j \0 \CheckNs
\advance \k \1 \FindPrimes
l.73 \FindPrimes
? X
</usr/local/texlive/2012/texmf-dist/fonts/type1/public/amsfonts/cm/cmr10.pfb>
Output written on primes.pdf (150471 pages, 28821382 bytes).
Transcript written on primes.log.
The file primes.pdf is not corrupted here are the last pages for this run:

Ah, and some confirmation:
|\^/| Maple 12 (IBM INTEL LINUX)
._|\| |/|_. Copyright (c) Maplesoft, a division of Waterloo Maple Inc. 2008
\ MAPLE / All rights reserved. Maple is a trademark of
<____ ____> Waterloo Maple Inc.
| Type ? for help.
> ithprime (150471);
2021959
Also, the code writes the primes into the log file: (I don't know the origin of some additional linebreaks in the log file, like here just above page number 150467)
] [150461
2021807
] [150462
2021837
] [150463
2021839
] [150464
2021843
] [150465
2021849
] [150466
2021863
]
[150467
2021879
] [150468
2021891
] [150469
2021927
] [150470
2021933
] [150471
2021959
]
! Interruption.
Here is the code (just the basic algorithm of testing for divisors until square root of n, without adornments; and minus an unfortunate oversight of the previous version, as explained above):
% Time-stamp: <30-01-2013 19:19:16 CET jfbu>
% file primes.tex to find and count prime numbers
% page geometry
\pdfoutput = 1
\hoffset = -2.44cm
\voffset = -2.44cm
\pdfpagewidth = 7cm
\pdfpageheight = .7cm
\hsize 6.8cm
\vsize .5cm
\parindent 0pt
\footline {}
\chardef\zero 0
\chardef\one 1
\chardef\two 2
\newcount\mone
\mone = -1
\newcount\n
\newcount\m
\newcount\e
\newcount\k
\newcount\j
\def\TestDiv{%
\advance\e\two
\ifnum\e>\k\else
\m\n
\divide\m\e
\multiply\m\e
\ifnum\m=\n \else
\expandafter\expandafter\expandafter
\TestDiv
\fi
\fi}
\def\ToLogIfPrime{\e\one
\TestDiv
\ifnum\e>\k
% \write\sixteen{\the\n}\leavevmode\vfill\eject
\write\mone{\the\n}\the\n\ (\the\count\zero)\vfill\eject
\fi}
\def\CheckNs{%
\ifnum\j<\k
\advance\n\two \ToLogIfPrime
\advance\j\one
\expandafter\CheckNs
\fi}
\def\FindPrimes{%
\j\zero % k is even,
\CheckNs % we scan the k odd n's such that k^2 < n < (k+1)^2
\advance\k\one % now k is odd and the next odd n is k^2
\advance\n\two % which certainly is not prime.
\j\zero
\CheckNs % from k^2+2 up to (k+1)^2 - 1, this makes k n's to go
%\ifnum\k<100 % <- set this if you want finite computation
\advance\k\one
%\expandafter
\FindPrimes
%\fi
}
\k 2
\n 3
\write-1{2}2 (1)\vfill\eject
\write-1{3}3 (2)\vfill\eject
%
\FindPrimes
\bye
\def\foo{\foo}\foo? – cgnieder Jan 29 '13 at 00:38\count255=0 \loop\ifnum\count255>0 Print something\endgraf\repeat– egreg Jan 29 '13 at 10:44foreverprimitive loop. – egreg Jan 30 '13 at 10:50