Mike wrote:
> The read and write work ok if I use a 5ms after the write,
> but I would like to use polling for speed.
>
> I send <start> 10100000 then check for an acknowledge,
> but it doesn't seem to work.

That's right.
They go 'off line' while busy writing,
by not responding to their address.
This is easier than polling a status bit inside the EEPROM
- there are none to my knowledge.

What you need to do is get your master to retry
sending the I2C message if the slave does not acknowledge.

When the EEPROM is ready, it will acknowledge its address,
and accept the message.

Read the data sheets.