"Interdum sane, qui cogitare se dicit nondum constitutum sibi certi quidquam fatetur"

What follows is not an answer but just contains some (grammatically relevant) remarks on the translation given above by brianpck. I've decided to include these remarks here as an answer since I'm afraid I've already posted too many comments in brianpck's post...

Let's start with the following fragment:

nondum constitutum sibi "This is a dative of agent (thanks to TKR for the correction), meaning 'not yet decided by him.'"

This translation could set some learners of Latin like d_e wondering: "fine, but then is there any semantic difference between a "dative of agent" (sibi) and an "ablative of agent" (a se)? Should both be translated as "by him"?

When dealing with the difference between "datives of agent" and "ablatives of agent", some (good) grammars of Latin often include relevant pairs like the following one:

Mihi consilium captum diu est (Cic. Fam. 5.19) 'lit. 'For me there is a decision taken for long time' (NB I: the participle captum is NOT part of the verb(al form) but is rather acting as an adjectival predicative of the nominative subject consilium; NB II: for reasons clarified below, the existential "For me there is..." is better rendered in many languages via a possesive verb: "I have + Direct Object + predicative of DO").

A me consilium captum est 'The decision {has been/was taken} by me' (NB: the participle captum is part of the complex verbal form).

Cf. the following relevant remark made by the French latinist Lavency (1985: 165), which, I hope, is quite transparent for English speakers (otherwise, please let me know): "on notera la différence entre Mihi (Dat.) consilium captum est (présent correspondant à: Mihi consilium captum fuit) et A me (ab + Abl.) consilium captum est (passé correspondant à: A me consilium capitur)". Importantly, note that the participle captum is adjectival in the former example with mihi, whereas it is verbal in the latter with a me (for further discussion on the important distinction between "adjectival passives" and "verbal passives", see this link).

As I've pointed out in some comments on brianpck's answer, the notion of possession is relevant when dealing with so-called "datives of agent". For example, authors of excellent descriptive grammars of Latin like Lavency (1985: 165-166) have noted parallelisms like the one between Caesari equitatus coactus erat and Caesar equitatum coactum habebat (Caes. BG 1.15). Indeed, there is a close relationship between so-called "datives of agent" and "datives of possession". So, for the purposes of the present post, note that a parallelism can also be put forward between the following constructions. The former involves a dative of agent sibi (which is in fact a dative of possession!), whereas the latter involves the possessive verb habere:

nondum constitutum (esse) sibi certi quidquam

nondum constitutum habere se certi quidquam

To recapitulate, consider the full translation given by brianpck:

Sometimes, of course, he who says that he is thinking [cogitare] is claiming that nothing certain has been decided by him yet.

Note that this translation has transformed:

(i) a Latin adjectival passive INTO an English verbal passive (please note that in Classical Latin the adjectival participle constitutum, when coappearing, for example, with a "dative of agent", does NOT form a perfect infinitive with esse);

(ii) a Latin dative of agent (a dative of possession; see above) INTO an English by-phrase (the latter being more similar to the Latin "ablative of agent").

In short, note that brianpck's translation is rather a translation of nondum constitutum a se (certi quidquam), where constitutum does form a perfect infinitive with an elliptical esse.

Please don't get me wrong! I'm NOT saying that there are some "mistakes" in brianpck's translation. The aim of the present "answer" is just to exemplify how some subtle grammatical facts of the original text can change when translating them to another language. As is well-known, translations often involve, i.a., a non-trivial process of (natural) adaptation of the original information to the target language, whereby some subtle nuances in the original text can be lost. To rephrase/reuse an admittedly polemical title (I know some of you strongly disagree with Newmeyer's statement!), "translation is translation and grammar is grammar".

Take-home message: As a Spanish colleague of mine (Ignacio Bosque, a famous grammarian of this language) likes to say, "en la lengua —y en particular en la Gramática— todo son matices" ('in language -and in particular in Grammar- everything is a matter of nuances').

FURTHER READING:

• For related discussion on the different grammatical behavior (e.g., the different syntactic distribution) of "datives of agent" and "ablatives of agent", see this link.
• Take a look at this excellent descriptive article on datives of agent (NB: it is not technical at all but is written in French).

Here's how I would translate the bolded phrase:

Sometimes, of course, he who says that he is thinking [cogitare] is claiming that nothing certain has been decided by him yet.

Here's a breakdown:

qui cogitare se dicit...

This is the subject: "He who says that he is thinking"

fatetur

This is the verb: "is claiming"

nondum constitutum sibi

This is a dative of agent (thanks to TKR for the correction), meaning "not yet decided by him."

certi quidquam

The certi is an example of a partitive genitive, e.g. "quid novi" = "something new." In this case, "quidquam certi" means "something certain." There is an understood "esse" that links this to the previous: "nothing certain has been decided by him yet."